├── .gitignore ├── LICENSE ├── Makefile ├── README.md ├── alignment ├── Makefile └── identify_chunk.c ├── crowded ├── ConsumerProducerQueue.h ├── Makefile ├── chunk_multi_thread_aio └── multi_thread_aio_for_chunk.cpp ├── fio ├── fill_all.fio ├── fio-mix.fio.template ├── fio-randread.fio.template ├── fio-randwrite.fio.template ├── fio-seqread.fio.template ├── fio-seqwrite.fio.template └── trim_all.fio ├── identify_chunk.c ├── identify_mapping.cpp ├── init.c ├── interleaving ├── ConsumerProducerQueue.h ├── Makefile ├── multi_thread_aio └── multi_thread_aio.cpp ├── mapping ├── Makefile ├── identify_mapping.cpp └── random_overwrite.cpp ├── multi_thread_aio.cpp ├── multi_thread_aio_for_pair.cpp ├── random_overwrite ├── random_overwrite.cpp ├── setup_cpu.sh └── trim_all.fio /.gitignore: -------------------------------------------------------------------------------- 1 | # Prerequisites 2 | *.d 3 | 4 | # Compiled Object files 5 | *.slo 6 | *.lo 7 | *.o 8 | *.obj 9 | 10 | # Precompiled Headers 11 | *.gch 12 | *.pch 13 | 14 | # Compiled Dynamic libraries 15 | *.so 16 | *.dylib 17 | *.dll 18 | 19 | # Fortran module files 20 | *.mod 21 | *.smod 22 | 23 | # Compiled Static libraries 24 | *.lai 25 | *.la 26 | *.a 27 | *.lib 28 | 29 | # Executables 30 | *.exe 31 | *.out 32 | *.app 33 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | 635 | Copyright (C) 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | Copyright (C) 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /Makefile: -------------------------------------------------------------------------------- 1 | test: raw_test.c 2 | gcc -o raw_test raw_test.c 3 | 4 | init: init.c 5 | gcc -o init -D_GNU_SOURCE init.c 6 | 7 | 8 | chunk: identify_chunk.c 9 | gcc -o identify_chunk -D_GNU_SOURCE identify_chunk.c 10 | 11 | interleave: identify_interleave.cpp 12 | g++ -std=c++11 -o identify_interleave -D_GNU_SOURCE identify_interleave.cpp -lpthread 13 | 14 | mapping: identify_mapping.cpp 15 | g++ -std=c++11 -o identify_mapping -D_GNU_SOURCE identify_mapping.cpp -lpthread 16 | 17 | split_request: split_request.cpp 18 | g++ -std=c++11 -o split_request -D_GNU_SOURCE split_request.cpp -lpthread 19 | 20 | shear_pattern: shear_identify_pattern.cpp 21 | g++ -std=c++11 -o shear_pattern -D_GNU_SOURCE shear_identify_pattern.cpp -lpthread 22 | 23 | random_overwrite: random_overwrite.cpp 24 | g++ -std=c++11 -o random_overwrite -D_GNU_SOURCE random_overwrite.cpp -lpthread 25 | 26 | multi_thread_aio: multi_thread_aio.cpp ConsumerProducerQueue.h 27 | g++ -std=c++11 -o multi_thread_aio -D_GNU_SOURCE multi_thread_aio.cpp ConsumerProducerQueue.h -lpthread -laio 28 | 29 | chunk_multi_thread_aio: multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h 30 | g++ -std=c++11 -o chunk_multi_thread_aio -D_GNU_SOURCE multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h -lpthread -laio 31 | 32 | pair_read: pair_read.cpp 33 | g++ -std=c++11 -o pair_read -D_GNU_SOURCE pair_read.cpp -lpthread 34 | 35 | clean: 36 | rm init raw_test identify_chunk identify_interleave identify_mapping 37 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # OptaneBench 2 | Benchmark for Intel Optane SSD, towards an unwritten contract of Intel Optane SSD 3 | 4 | 5 | # Microworkloads 6 | 7 | ## Random/Sequential workloads: (Figure 1, 4, 5, 7, 8, 10) 8 | We control workloads with FIO. variables mainly include 0) I/O pattern 2) Queue depth and 3) request size. 9 | rand read template: fio/fio-randread.fio.template 10 | sequential read template: fio/fio-seqread.fio.template 11 | 12 | 13 | trim\_all.fio before each write workload (to avoid influence from garbage collection) 14 | rand write template: fio/fio-randread.fio.template 15 | sequential write template: fio/fio-seqread.fio.template 16 | 17 | To note: overall queue depth = numjobs * iodepth (we use numjobs = 4 when overall QD >= 4, numjobs = 1 when overall QD < 4) 18 | 19 | Mix: 20 | 1. First init 10% of space with fill\_all.fio 21 | 2. Use fio-mix.fio.template to issue mixed workloads 22 | 23 | ## Detecting Interleaving Degree (Figure 3) 24 | * reference: HPDC 25 | * inititialize the device: 26 | * trim all data: fio trim\_all.fio 27 | * fill the storage space using sequential writes with a request size of 256KB: ./init device\_name device\_size(in sector) 512 28 | * test with different stride: 29 | * ./multi\_thread\_aio /dev/nvme0n1 8 2 8 jump 16 30 | * dev\_name D(offset to jump, in unit of sector) j(num\_parallel\_jobs to submit IO) d(stride/io\_size in sector) read\_type(random/seq/jump) queue\_depth 31 | 32 | ## Parallel Sectors (Figure 6) 33 | * ./chunk\_multi\_thread\_aio /dev/nvme0n1 0 2 1 randomread 4 34 | * chunk\_multi\_thread\_aio dev\_name D(not useful) j(num\_parallel\_jobs to submit IO) d(io\_size in sector) read\_type(random/seq/jump) queue\_depth(how many parallel sector reads to a 4KB chunk) 35 | * blktrace is recommended to track latency of each request 36 | 37 | ## Identifying Alignment Influence (Figure 9) 38 | * reference: HPDC 39 | * identify\_chunk dev\_name M(Estimated Max chunk size) n(req\_size) k(offset) 40 | * M = 64 during the test. vary n and k to test (both in unit of sector) 41 | 42 | ## Detecting Mapping Policy (Figure 11) 43 | * init with different written order 44 | * init: sequential write 45 | * random write: random\_overwrite dev\_name S(overwritten zone in GB, start from 0) n(req\_size in sector) 46 | * read according to different order: 47 | * identify\_mapping dev\_name D(Estimated Max interleave degree) j(num\_parallel\_jobs) d(stride in chunk(16 sectors)) write\_order(write order file name) 48 | 49 | -------------------------------------------------------------------------------- /alignment/Makefile: -------------------------------------------------------------------------------- 1 | test: raw_test.c 2 | gcc -o raw_test raw_test.c 3 | 4 | init: init.c 5 | gcc -o init -D_GNU_SOURCE init.c 6 | 7 | 8 | chunk: identify_chunk.c 9 | gcc -o identify_chunk -D_GNU_SOURCE identify_chunk.c 10 | 11 | interleave: identify_interleave.cpp 12 | g++ -std=c++11 -o identify_interleave -D_GNU_SOURCE identify_interleave.cpp -lpthread 13 | 14 | mapping: identify_mapping.cpp 15 | g++ -std=c++11 -o identify_mapping -D_GNU_SOURCE identify_mapping.cpp -lpthread 16 | 17 | split_request: split_request.cpp 18 | g++ -std=c++11 -o split_request -D_GNU_SOURCE split_request.cpp -lpthread 19 | 20 | shear_pattern: shear_identify_pattern.cpp 21 | g++ -std=c++11 -o shear_pattern -D_GNU_SOURCE shear_identify_pattern.cpp -lpthread 22 | 23 | random_overwrite: random_overwrite.cpp 24 | g++ -std=c++11 -o random_overwrite -D_GNU_SOURCE random_overwrite.cpp -lpthread 25 | 26 | multi_thread_aio: multi_thread_aio.cpp ConsumerProducerQueue.h 27 | g++ -std=c++11 -o multi_thread_aio -D_GNU_SOURCE multi_thread_aio.cpp ConsumerProducerQueue.h -lpthread -laio 28 | 29 | chunk_multi_thread_aio: multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h 30 | g++ -std=c++11 -o chunk_multi_thread_aio -D_GNU_SOURCE multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h -lpthread -laio 31 | 32 | pair_read: pair_read.cpp 33 | g++ -std=c++11 -o pair_read -D_GNU_SOURCE pair_read.cpp -lpthread 34 | 35 | clean: 36 | rm init raw_test identify_chunk identify_interleave identify_mapping 37 | -------------------------------------------------------------------------------- /alignment/identify_chunk.c: -------------------------------------------------------------------------------- 1 | #define _GNU_SOURCE 2 | 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | 12 | // Note: all pos in sector unless before passed in read/write 13 | int SECTOR_SIZE = 512; 14 | 15 | long rand_pos(int Align) { 16 | //return Align * rand(8*1024*1024*2/Align); // randomly span 8GB, while aligned to Align 17 | return Align * (rand() %(100*1024*1024*2/Align)); // randomly span 8/200GB, while aligned to Align 18 | } 19 | 20 | int main(int argc, char* * argv) { 21 | printf("===== Identify Chunk Size of Specified Device =====\n"); 22 | // identify chunk size 23 | if (argc < 5) { 24 | printf("Wrong parameters: identify_chunk dev_name M(Estimated Max chunk size) n(req_size) k(offset)\n"); 25 | return 1; 26 | } 27 | int M = atoi(argv[2]); 28 | int n = atoi(argv[3]); 29 | int k = atoi(argv[4]); 30 | char * read_buf = (char *) malloc(sizeof(char) * (n * SECTOR_SIZE + SECTOR_SIZE)); 31 | int ret = posix_memalign((void **)&read_buf, SECTOR_SIZE, n*SECTOR_SIZE); 32 | 33 | printf("To run with:\n aligned_chunk: %d, request_size: %d, offset: %d, device_name: %s\n", M, n, k, argv[1]); 34 | 35 | // open raw block device 36 | int fd = open(argv[1], O_RDONLY | O_DIRECT); // O_DIRECT????? 37 | if (fd < 0) { 38 | printf("Raw Device Open failed\n"); 39 | return 1; 40 | } 41 | srand(time(0)); 42 | 43 | for (int i = 0; i < 500000; i++) { 44 | long pos = rand_pos(M) + k; 45 | //printf("Random Position: %ld \n", pos * SECTOR_SIZE); 46 | int sz = pread(fd, read_buf, n * SECTOR_SIZE, pos * SECTOR_SIZE); 47 | //int sz = read(fd, read_buf, n * SECTOR_SIZE); 48 | assert(sz == n*SECTOR_SIZE); 49 | //printf("Read in %d bytes\n", sz); 50 | } 51 | 52 | printf("Experiment done\n"); 53 | close(fd); 54 | return 0; 55 | } 56 | -------------------------------------------------------------------------------- /crowded/ConsumerProducerQueue.h: -------------------------------------------------------------------------------- 1 | #ifndef __CONSUMERPRODUCERQUEUE_H__ 2 | #define __CONSUMERPRODUCERQUEUE_H__ 3 | 4 | #include 5 | #include 6 | #include 7 | 8 | /* 9 | * Some references in order 10 | * 11 | * Some code I wrote a long time before C++ 11 to do consumer producer buffers, using 2 condition variables 12 | * https://github.com/mdaus/coda-oss/blob/master/modules/c%2B%2B/mt/include/mt/RequestQueue.h 13 | * 14 | * A great article explaining both 2 condition variable and 1 condition variable buffers 15 | * https://en.wikipedia.org/wiki/Monitor_%28synchronization%29#Condition_variables 16 | * 17 | * C++ 11 thread reference: 18 | * http://en.cppreference.com/w/cpp/thread 19 | */ 20 | template 21 | class ConsumerProducerQueue 22 | { 23 | std::condition_variable cond; 24 | std::mutex mutex; 25 | std::queue cpq; 26 | int maxSize; 27 | public: 28 | ConsumerProducerQueue() 29 | { } 30 | 31 | void set_max(int max) { 32 | maxSize = max; 33 | } 34 | void add(T request) 35 | { 36 | std::unique_lock lock(mutex); 37 | cond.wait(lock, [this]() 38 | { return !isFull(); }); 39 | cpq.push(request); 40 | lock.unlock(); 41 | cond.notify_all(); 42 | } 43 | 44 | void consume(T &request) 45 | { 46 | std::unique_lock lock(mutex); 47 | cond.wait(lock, [this]() 48 | { return !isEmpty(); }); 49 | request = cpq.front(); 50 | cpq.pop(); 51 | lock.unlock(); 52 | cond.notify_all(); 53 | 54 | } 55 | 56 | bool isFull() const 57 | { 58 | return cpq.size() >= maxSize; 59 | } 60 | 61 | bool isEmpty() const 62 | { 63 | return cpq.size() == 0; 64 | } 65 | 66 | int length() const 67 | { 68 | return cpq.size(); 69 | } 70 | 71 | void clear() 72 | { 73 | std::unique_lock lock(mutex); 74 | while (!isEmpty()) 75 | { 76 | cpq.pop(); 77 | } 78 | lock.unlock(); 79 | cond.notify_all(); 80 | } 81 | }; 82 | 83 | #endif 84 | -------------------------------------------------------------------------------- /crowded/Makefile: -------------------------------------------------------------------------------- 1 | test: raw_test.c 2 | gcc -o raw_test raw_test.c 3 | 4 | init: init.c 5 | gcc -o init -D_GNU_SOURCE init.c 6 | 7 | 8 | chunk: identify_chunk.c 9 | gcc -o identify_chunk -D_GNU_SOURCE identify_chunk.c 10 | 11 | interleave: identify_interleave.cpp 12 | g++ -std=c++11 -o identify_interleave -D_GNU_SOURCE identify_interleave.cpp -lpthread 13 | 14 | mapping: identify_mapping.cpp 15 | g++ -std=c++11 -o identify_mapping -D_GNU_SOURCE identify_mapping.cpp -lpthread 16 | 17 | split_request: split_request.cpp 18 | g++ -std=c++11 -o split_request -D_GNU_SOURCE split_request.cpp -lpthread 19 | 20 | shear_pattern: shear_identify_pattern.cpp 21 | g++ -std=c++11 -o shear_pattern -D_GNU_SOURCE shear_identify_pattern.cpp -lpthread 22 | 23 | random_overwrite: random_overwrite.cpp 24 | g++ -std=c++11 -o random_overwrite -D_GNU_SOURCE random_overwrite.cpp -lpthread 25 | 26 | multi_thread_aio: multi_thread_aio.cpp ConsumerProducerQueue.h 27 | g++ -std=c++11 -o multi_thread_aio -D_GNU_SOURCE multi_thread_aio.cpp ConsumerProducerQueue.h -lpthread -laio 28 | 29 | chunk_multi_thread_aio: multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h 30 | g++ -std=c++11 -o chunk_multi_thread_aio -D_GNU_SOURCE multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h -lpthread -laio 31 | 32 | pair_read: pair_read.cpp 33 | g++ -std=c++11 -o pair_read -D_GNU_SOURCE pair_read.cpp -lpthread 34 | 35 | clean: 36 | rm init raw_test identify_chunk identify_interleave identify_mapping 37 | -------------------------------------------------------------------------------- /crowded/chunk_multi_thread_aio: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/sherlockwu/OptaneBench/90ef168cbd79cbfbc57122069b7cd79474bd325c/crowded/chunk_multi_thread_aio -------------------------------------------------------------------------------- /crowded/multi_thread_aio_for_chunk.cpp: -------------------------------------------------------------------------------- 1 | #define _GNU_SOURCE 2 | 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | #include 12 | #include 13 | #include 14 | #include 15 | #include 16 | #include 17 | #include 18 | #include 19 | #include 20 | #include "ConsumerProducerQueue.h" 21 | 22 | #define MAX_COUNT 65536 23 | 24 | std::atomic pointer(0); 25 | std::vector read_order; 26 | 27 | // Note: all pos in sector unless before passed in read/write 28 | int SECTOR_SIZE = 512; 29 | int STRIDE_SIZE = SECTOR_SIZE * 1; // chunk size actually 30 | int num_ios = 100000; 31 | int completed_ios = 0; 32 | 33 | int D = 0; 34 | int j = 0; 35 | int d = 0; 36 | int fd = 0; 37 | io_context_t ctx_; 38 | int TRIES = 5; // for shear 39 | long overall_us = 0; 40 | 41 | std::condition_variable cond; 42 | std::mutex mutex; 43 | int queue_depth = 0; 44 | int max_qd = 16; 45 | ConsumerProducerQueue job_queue; 46 | int pattern = 7; 47 | 48 | struct io_event events[MAX_COUNT]; 49 | struct timespec timeout; 50 | char * read_buf; 51 | 52 | void *eachThread(void *vargp) 53 | { 54 | // consumer 55 | int id = *(int*)vargp; 56 | long pos; // in unit of sector 57 | printf("Thread %d ready to run \n", id); 58 | 59 | int sz; 60 | 61 | for ( ; ; ) { 62 | // Consumer wait the queue depth control 63 | job_queue.consume(pos); 64 | if (pos == -1) { 65 | break; 66 | } 67 | //std::cout << "to issue IO " << pos/4096 << std::endl; 68 | struct iocb * p = (struct iocb *)malloc(sizeof(struct iocb)); 69 | //io_prep_pread(p, fd_[item.fd], item.buffer, item.length, item.offset); 70 | io_prep_pread(p, fd, read_buf, STRIDE_SIZE, pos); 71 | //p->data = (void *) item.io_status; 72 | p->data = (void *) read_buf; 73 | 74 | if (io_submit(ctx_, 1, &p) != 1) { 75 | io_destroy(ctx_); 76 | std::cout << "io submit error" << std::endl; 77 | exit(1); 78 | } 79 | } 80 | std::cout << "Thread " << id << " issued all IOs" << std::endl; 81 | return NULL; 82 | } 83 | 84 | 85 | void generate_read_trace(char type) { 86 | if (type == 's') { 87 | std::cout << "running shear: qd: " << max_qd << " segment size(in sector): " << D*d << std::endl; 88 | srand(time(0)); 89 | for (int i = 0; i < num_ios; i++) // generate random segments 90 | read_order.push_back((long)((rand() %(800*1024*1024*2/(D*d))))*D*STRIDE_SIZE); 91 | return; 92 | } else if (type == 'r') { 93 | std::cout << "To read randomly" << std::endl; 94 | srand(time(0)); 95 | for (int i = 0; i < num_ios; i++) 96 | read_order.push_back((long)((rand() %(800*1024*1024/4)))*4096); 97 | } else { 98 | std::cout << "Wrong choice of workloads, should be seq/random" << std::endl; 99 | exit(1); 100 | } 101 | 102 | return; 103 | } 104 | 105 | int main(int argc, char* * argv) { 106 | printf("===== Multi-thread libaio to Specified Device =====\n"); 107 | // identify chunk size 108 | if (argc < 7) { 109 | printf("Wrong parameters: multi_thread_aio dev_name D(offset to jump, in unit of sector) j(num_parallel_jobs to submit IO) d(stride/io_size in sector) read_type(random/seq/jump) queue_depth\n"); 110 | return 1; 111 | } 112 | D = atoi(argv[2]); //offset between two reads, in unit of sector 113 | j = atoi(argv[3]); 114 | d = atoi(argv[4]); // request size 115 | STRIDE_SIZE = SECTOR_SIZE * d; 116 | max_qd = atoi(argv[6]); 117 | int FIRST = atoi(argv[7]); 118 | //max_qd = 16; 119 | pattern = atoi(argv[6])/16; 120 | job_queue.set_max(max_qd); 121 | 122 | printf("To run with:\n jump: %d, num_jobs: %d, stride: %d, device_name: %s, write_order: %s, max_qd: %d\n", D, j, d, argv[1], argv[5], max_qd); 123 | 124 | // open raw block device 125 | fd = open(argv[1], O_RDONLY | O_DIRECT); // O_DIRECT 126 | if (fd < 0) { 127 | printf("Raw Device Open failed\n"); 128 | return 1; 129 | } 130 | // context to do async IO 131 | memset(&ctx_, 0, sizeof(ctx_)); 132 | if (io_setup(MAX_COUNT, &ctx_) != 0) { 133 | std::cout << "io_context_t set failed" << std::endl; 134 | exit(1); 135 | } 136 | // generate read trace 137 | // 138 | generate_read_trace(argv[5][0]); 139 | 140 | //parallel reads 141 | 142 | // start the number of threads 143 | pthread_t * thread_pool = (pthread_t *)malloc(sizeof(pthread_t) * j); 144 | 145 | 146 | for (int i = 0; i < j; i++) { 147 | int *arg = (int *)malloc(sizeof(*arg)); 148 | *arg = i; 149 | pthread_create(&thread_pool[i], NULL, eachThread, (void *)arg); 150 | } 151 | 152 | 153 | read_buf = (char *) malloc(sizeof(char) * (STRIDE_SIZE + SECTOR_SIZE)); 154 | int ret = posix_memalign((void **)&read_buf, SECTOR_SIZE, STRIDE_SIZE + SECTOR_SIZE); 155 | sleep(1); 156 | 157 | //as a producer 158 | struct timeval start, end; 159 | gettimeofday(&start, NULL); 160 | int onflight_io = 0; 161 | 162 | timeout.tv_sec = 0; 163 | timeout.tv_nsec = 100; 164 | for (int i = 0; i < num_ios; i++) { 165 | // for each request 166 | onflight_io = 0; 167 | if (argv[5][0] == 'r') { 168 | // parallel reads 169 | for (int k = 0; k < max_qd; k++) { 170 | job_queue.add(read_order[i] + k*SECTOR_SIZE); 171 | } 172 | onflight_io = max_qd; 173 | } else if (argv[5][0] == 'p') { 174 | job_queue.add(FIRST*STRIDE_SIZE); 175 | job_queue.add((FIRST+1)*STRIDE_SIZE + D*SECTOR_SIZE); 176 | onflight_io = 2; 177 | } 178 | // wait until all finished 179 | while (onflight_io > 0) { 180 | int ret = io_getevents(ctx_, 0, MAX_COUNT, events, &timeout); 181 | if (ret < 0) { 182 | std::cout << "Getevents Error" << std::endl; 183 | exit(1); 184 | } 185 | if (ret > 0) { 186 | onflight_io -= ret; 187 | } 188 | } 189 | } 190 | 191 | //wait until all IO finished 192 | while (onflight_io) { 193 | int ret = io_getevents(ctx_, 0, MAX_COUNT, events, &timeout); 194 | if (ret < 0) { 195 | std::cout << "Getevents Error" << std::endl; 196 | exit(1); 197 | } 198 | if (ret > 0) { 199 | onflight_io -= ret; 200 | //std::cout << "=== Get Events("<< ret << "): " << ret << std::endl; 201 | //for (int j = 0; j < ret; j++) { 202 | // std::cout << "=== Complete " << "::" << events[j].obj->aio_fildes << std::endl; 203 | //} 204 | } 205 | } 206 | 207 | 208 | 209 | gettimeofday(&end, NULL); 210 | long time_us = ((end.tv_sec * 1000000 + end.tv_usec) 211 | - (start.tv_sec * 1000000 + start.tv_usec)); 212 | 213 | overall_us += time_us; 214 | // Signal to exit 215 | for (int i = 0; i < j; i++) { 216 | job_queue.add(-1); 217 | } 218 | for (int i = 0; i < j; i++) { 219 | pthread_join(thread_pool[i], NULL); 220 | } 221 | 222 | std::cout << "All IO finished" << std::endl; 223 | printf("Time taken: %ld us, Bandwidth: %f MB/s \n", time_us, (float)((long)STRIDE_SIZE*num_ios)/1024/1024*1000000/time_us); 224 | //printf("Time taken: %ld us, Bandwidth: %f MB/s \n", overall_us/TRIES, (float)((long)STRIDE_SIZE*num_ios)/1024/1024*1000000/(overall_us/TRIES)); 225 | 226 | free(read_buf); 227 | close(fd); 228 | return 0; 229 | } 230 | -------------------------------------------------------------------------------- /fio/fill_all.fio: -------------------------------------------------------------------------------- 1 | [global] 2 | name=fio-fill 3 | ioengine=libaio 4 | direct=1 5 | thread=1 6 | buffered=0 7 | randrepeat=0 8 | ramp_time=0 9 | norandommap 10 | refill_buffers 11 | log_max_value=1 12 | group_reporting 13 | percentile_list=50.0:99.0 14 | filename=/dev/nvme1n1 15 | 16 | [preconditioning] 17 | stonewall 18 | size=10% 19 | bs=256k 20 | iodepth=16 21 | numjobs=1 22 | rw=write 23 | 24 | -------------------------------------------------------------------------------- /fio/fio-mix.fio.template: -------------------------------------------------------------------------------- 1 | ; fio-rand-write.job for fiotest 2 | 3 | [global] 4 | name=fio-mix 5 | ioengine=libaio 6 | direct=1 7 | thread=1 8 | buffered=0 9 | size=10% 10 | randrepeat=0 11 | time_based 12 | ramp_time=0 13 | norandommap 14 | refill_buffers 15 | log_max_value=1 16 | group_reporting 17 | percentile_list=25.0:50.0:75.0:99.0 18 | filename=/dev/nvme1n1 19 | 20 | 21 | [file1] 22 | bs=4k 23 | iodepth=1 24 | numjobs=4 25 | rw=randrw 26 | rwmixwrite=100 27 | runtime=20 28 | -------------------------------------------------------------------------------- /fio/fio-randread.fio.template: -------------------------------------------------------------------------------- 1 | ; fio-rand-read.job for fiotest 2 | 3 | [global] 4 | name=fio-rand-read 5 | ioengine=libaio 6 | direct=1 7 | thread=1 8 | buffered=0 9 | size=100% 10 | randrepeat=0 11 | time_based 12 | ramp_time=0 13 | norandommap 14 | refill_buffers 15 | log_max_value=1 16 | group_reporting 17 | percentile_list=25.0:50.0:75.0:99.0 18 | filename=/dev/nvme0n1 19 | 20 | 21 | [file1] 22 | ;stonewall 23 | bs=4k 24 | iodepth=1 25 | numjobs=4 26 | rw=randread 27 | runtime=20 28 | 29 | -------------------------------------------------------------------------------- /fio/fio-randwrite.fio.template: -------------------------------------------------------------------------------- 1 | ; fio-rand-write.job for fiotest 2 | 3 | [global] 4 | name=fio-rand-write 5 | ioengine=libaio 6 | direct=1 7 | thread=1 8 | buffered=0 9 | randrepeat=0 10 | ramp_time=0 11 | norandommap 12 | refill_buffers 13 | log_max_value=1 14 | group_reporting 15 | percentile_list=25.0:50.0:75.0:99.0 16 | filename=/dev/nvme1n1 17 | 18 | 19 | [file1] 20 | ;stonewall 21 | bs=4k 22 | iodepth=1 23 | numjobs=4 24 | rw=randwrite 25 | time_based 26 | runtime=20 27 | -------------------------------------------------------------------------------- /fio/fio-seqread.fio.template: -------------------------------------------------------------------------------- 1 | ; fio-rand-read.job for fiotest 2 | 3 | [global] 4 | name=fio-rand-read 5 | ioengine=libaio 6 | direct=1 7 | thread=1 8 | buffered=0 9 | size=100% 10 | randrepeat=0 11 | time_based 12 | ramp_time=0 13 | norandommap 14 | refill_buffers 15 | offset_increment=150g 16 | log_max_value=1 17 | group_reporting 18 | percentile_list=50.0:99.0 19 | filename=/dev/nvme0n1 20 | 21 | 22 | [stream1] 23 | ;stonewall 24 | bs=4k 25 | iodepth=1 26 | numjobs=1 27 | rw=read 28 | runtime=20 29 | 30 | -------------------------------------------------------------------------------- /fio/fio-seqwrite.fio.template: -------------------------------------------------------------------------------- 1 | ; fio-seqwrite.job for fiotest 2 | 3 | [global] 4 | name=fio-seq-write 5 | ioengine=libaio 6 | direct=1 7 | thread=1 8 | buffered=0 9 | randrepeat=0 10 | ramp_time=0 11 | norandommap 12 | refill_buffers 13 | offset_increment=100g 14 | log_max_value=1 15 | group_reporting 16 | percentile_list=50.0:99.0 17 | filename=/dev/nvme1n1 18 | 19 | 20 | [stream1] 21 | bs=4k 22 | iodepth=1 23 | numjobs=1 24 | rw=write 25 | time_based 26 | runtime=20 27 | -------------------------------------------------------------------------------- /fio/trim_all.fio: -------------------------------------------------------------------------------- 1 | ; trim all data 2 | 3 | [global] 4 | name=fio-trim 5 | ioengine=libaio 6 | direct=1 7 | thread=1 8 | buffered=0 9 | randrepeat=0 10 | ramp_time=0 11 | norandommap 12 | refill_buffers 13 | log_max_value=1 14 | group_reporting 15 | percentile_list=50.0:99.0 16 | filename=/dev/nvme1n1 17 | 18 | [preconditioning] 19 | stonewall 20 | size=100% 21 | bs=4m 22 | iodepth=16 23 | numjobs=1 24 | rw=trim 25 | 26 | -------------------------------------------------------------------------------- /identify_chunk.c: -------------------------------------------------------------------------------- 1 | #define _GNU_SOURCE 2 | 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | 12 | // Note: all pos in sector unless before passed in read/write 13 | int SECTOR_SIZE = 512; 14 | 15 | long rand_pos(int Align) { 16 | //return Align * rand(8*1024*1024*2/Align); // randomly span 8GB, while aligned to Align 17 | return Align * (rand() %(100*1024*1024*2/Align)); // randomly span 8/200GB, while aligned to Align 18 | } 19 | 20 | int main(int argc, char* * argv) { 21 | printf("===== Identify Chunk Size of Specified Device =====\n"); 22 | // identify chunk size 23 | if (argc < 5) { 24 | printf("Wrong parameters: identify_chunk dev_name M(Estimated Max chunk size) n(req_size) k(offset)\n"); 25 | return 1; 26 | } 27 | int M = atoi(argv[2]); 28 | int n = atoi(argv[3]); 29 | int k = atoi(argv[4]); 30 | char * read_buf = (char *) malloc(sizeof(char) * (n * SECTOR_SIZE + SECTOR_SIZE)); 31 | int ret = posix_memalign((void **)&read_buf, SECTOR_SIZE, n*SECTOR_SIZE); 32 | 33 | printf("To run with:\n aligned_chunk: %d, request_size: %d, offset: %d, device_name: %s\n", M, n, k, argv[1]); 34 | 35 | // open raw block device 36 | int fd = open(argv[1], O_RDONLY | O_DIRECT); // O_DIRECT????? 37 | if (fd < 0) { 38 | printf("Raw Device Open failed\n"); 39 | return 1; 40 | } 41 | srand(time(0)); 42 | 43 | for (int i = 0; i < 500000; i++) { 44 | long pos = rand_pos(M) + k; 45 | //printf("Random Position: %ld \n", pos * SECTOR_SIZE); 46 | int sz = pread(fd, read_buf, n * SECTOR_SIZE, pos * SECTOR_SIZE); 47 | //int sz = read(fd, read_buf, n * SECTOR_SIZE); 48 | assert(sz == n*SECTOR_SIZE); 49 | //printf("Read in %d bytes\n", sz); 50 | } 51 | 52 | printf("Experiment done\n"); 53 | close(fd); 54 | return 0; 55 | } 56 | -------------------------------------------------------------------------------- /identify_mapping.cpp: -------------------------------------------------------------------------------- 1 | #define _GNU_SOURCE 2 | 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | #include 12 | #include 13 | #include 14 | #include 15 | #include 16 | #include 17 | 18 | std::atomic pointer(0); 19 | std::vector write_order; 20 | 21 | // Note: all pos in sector unless before passed in read/write 22 | int SECTOR_SIZE = 512; 23 | int STRIDE_SIZE = SECTOR_SIZE * 16; // chunk size actually 24 | int num_ios = 10000; 25 | 26 | int jobs = 0; 27 | int j = 0; 28 | int d = 0; 29 | int fd = 0; 30 | void *eachThread(void *vargp) 31 | { 32 | int id = *(int*)vargp; 33 | long pos; // in unit of sector 34 | printf("Thread %d ready to run \n", id); 35 | char * read_buf = (char *) malloc(sizeof(char) * (STRIDE_SIZE + SECTOR_SIZE)); 36 | int ret = posix_memalign((void **)&read_buf, SECTOR_SIZE, STRIDE_SIZE + SECTOR_SIZE); 37 | 38 | int sz; 39 | while (jobs <= 0) ; 40 | 41 | // do IO 42 | 43 | for ( ; ; ) { 44 | pos = pointer.fetch_add(1); 45 | if (pos > num_ios) 46 | break; 47 | //printf("Position: %d \n", pos * d); 48 | //sz = pread(fd, read_buf, 8 * SECTOR_SIZE, (pos + i * big_stride)* SECTOR_SIZE); 49 | //sz = pread(fd, read_buf, 8 * SECTOR_SIZE, pointer * d * SECTOR_SIZE); 50 | //assert(sz == 8*SECTOR_SIZE); 51 | //sz = pread(fd, read_buf, STRIDE_SIZE, pos * d * STRIDE_SIZE); 52 | //sz = pread(fd, read_buf, STRIDE_SIZE, (pos/j) * 16 * STRIDE_SIZE + (pos%j) * d * STRIDE_SIZE); 53 | //sz = pread(fd, read_buf, STRIDE_SIZE, (pos%j) * d * STRIDE_SIZE); 54 | //sz = pread(fd, read_buf, STRIDE_SIZE, (pos%2) * d * STRIDE_SIZE); 55 | //sz = pread(fd, read_buf, STRIDE_SIZE, (pos%2) * d * STRIDE_SIZE); 56 | // according to write order 57 | //std::cout << pos << ":" << write_order[(pos%2) * d] << std::endl; 58 | //sz = pread(fd, read_buf, STRIDE_SIZE, write_order[(pos%2) * d]); 59 | sz = pread(fd, read_buf, STRIDE_SIZE, write_order[pos * (d+1)]); 60 | //printf("Done %ld %d\n", pointer*d*STRIDE_SIZE, sz); 61 | assert(sz == STRIDE_SIZE); 62 | } 63 | free(read_buf); 64 | return NULL; 65 | } 66 | 67 | 68 | 69 | int main(int argc, char* * argv) { 70 | printf("===== Identify Mapping of Specified Device(read according to writes order) =====\n"); 71 | // identify chunk size 72 | if (argc < 6) { 73 | printf("Wrong parameters: identify_mapping dev_name D(Estimated Max interleave degree) j(num_parallel_jobs) d(stride in chunk(16 sectors)) write_order(write order file name)\n"); 74 | return 1; 75 | } 76 | int D = atoi(argv[2]); 77 | j = atoi(argv[3]); 78 | d = atoi(argv[4]); 79 | 80 | printf("To run with:\n Max_Degree: %d, num_jobs: %d, stride: %d, device_name: %s, write_order: %s\n", D, j, d, argv[1], argv[5]); 81 | 82 | // open raw block device 83 | fd = open(argv[1], O_RDONLY | O_DIRECT); // O_DIRECT 84 | if (fd < 0) { 85 | printf("Raw Device Open failed\n"); 86 | return 1; 87 | } 88 | 89 | //parallel reads 90 | 91 | // start the number of threads 92 | pthread_t * thread_pool = (pthread_t *)malloc(sizeof(pthread_t) * j); 93 | 94 | for (int i = 0; i < j; i++) { 95 | int *arg = (int *)malloc(sizeof(*arg)); 96 | *arg = i; 97 | pthread_create(&thread_pool[i], NULL, eachThread, (void *)arg); 98 | } 99 | 100 | // prepare the read trace (according to pointer) 101 | std::ifstream write_trace(argv[5]); 102 | long cur_write; 103 | while(write_trace >> cur_write) { 104 | write_order.push_back(cur_write); 105 | } 106 | 107 | sleep(2); 108 | 109 | struct timeval start, end; 110 | gettimeofday(&start, NULL); 111 | jobs = j; // hint to start doing IOs 112 | for (int i = 0; i < j; i++) { 113 | pthread_join(thread_pool[i], NULL); 114 | } 115 | gettimeofday(&end, NULL); 116 | long time_us = ((end.tv_sec * 1000000 + end.tv_usec) 117 | - (start.tv_sec * 1000000 + start.tv_usec)); 118 | //printf("Time taken: %ld us, Bandwidth: %f MB/s \n", time_us, (float)(8*512*100000)/1024/1024*1000000/time_us); 119 | printf("Time taken: %ld us, Bandwidth: %f MB/s \n", time_us, (float)(STRIDE_SIZE*num_ios)/1024/1024*1000000/time_us); 120 | 121 | close(fd); 122 | return 0; 123 | } 124 | -------------------------------------------------------------------------------- /init.c: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | 10 | // Note: all pos in sector unless before passed in read/write 11 | int SECTOR_SIZE = 512; 12 | 13 | int main(int argc, char* * argv) { 14 | printf("===== Initialize Specified Device =====\n"); 15 | // identify chunk size 16 | if (argc < 4) { 17 | printf("Wrong parameters: init dev_name S(device_size in sector) n(req_size in sector) \n"); 18 | return 1; 19 | } 20 | int S = atoi(argv[2]); 21 | int n = atoi(argv[3]); 22 | 23 | char * write_buf = (char *) malloc(sizeof(char) * ((n+1) * SECTOR_SIZE)); 24 | int ret = posix_memalign((void **)&write_buf, SECTOR_SIZE, (n+1)*SECTOR_SIZE); 25 | 26 | printf("To run with:\n device_size: %d, request_size: %d, device_name: %s\n", S, n, argv[1]); 27 | 28 | // open raw block device 29 | int fd = open(argv[1], O_WRONLY | O_DIRECT); 30 | if (fd < 0) { 31 | printf("Raw Device Open failed\n"); 32 | return 1; 33 | } 34 | 35 | 36 | for (int i = 0; i < S; i+=n) { 37 | if (i%1000000 == 0) 38 | printf("Done %d\n", i); 39 | int sz = write(fd, write_buf, n * SECTOR_SIZE); 40 | assert(sz == n*SECTOR_SIZE); 41 | } 42 | 43 | free(write_buf); 44 | close(fd); 45 | return 0; 46 | } 47 | -------------------------------------------------------------------------------- /interleaving/ConsumerProducerQueue.h: -------------------------------------------------------------------------------- 1 | #ifndef __CONSUMERPRODUCERQUEUE_H__ 2 | #define __CONSUMERPRODUCERQUEUE_H__ 3 | 4 | #include 5 | #include 6 | #include 7 | 8 | /* 9 | * Some references in order 10 | * 11 | * Some code I wrote a long time before C++ 11 to do consumer producer buffers, using 2 condition variables 12 | * https://github.com/mdaus/coda-oss/blob/master/modules/c%2B%2B/mt/include/mt/RequestQueue.h 13 | * 14 | * A great article explaining both 2 condition variable and 1 condition variable buffers 15 | * https://en.wikipedia.org/wiki/Monitor_%28synchronization%29#Condition_variables 16 | * 17 | * C++ 11 thread reference: 18 | * http://en.cppreference.com/w/cpp/thread 19 | */ 20 | template 21 | class ConsumerProducerQueue 22 | { 23 | std::condition_variable cond; 24 | std::mutex mutex; 25 | std::queue cpq; 26 | int maxSize; 27 | public: 28 | ConsumerProducerQueue() 29 | { } 30 | 31 | void set_max(int max) { 32 | maxSize = max; 33 | } 34 | void add(T request) 35 | { 36 | std::unique_lock lock(mutex); 37 | cond.wait(lock, [this]() 38 | { return !isFull(); }); 39 | cpq.push(request); 40 | lock.unlock(); 41 | cond.notify_all(); 42 | } 43 | 44 | void consume(T &request) 45 | { 46 | std::unique_lock lock(mutex); 47 | cond.wait(lock, [this]() 48 | { return !isEmpty(); }); 49 | request = cpq.front(); 50 | cpq.pop(); 51 | lock.unlock(); 52 | cond.notify_all(); 53 | 54 | } 55 | 56 | bool isFull() const 57 | { 58 | return cpq.size() >= maxSize; 59 | } 60 | 61 | bool isEmpty() const 62 | { 63 | return cpq.size() == 0; 64 | } 65 | 66 | int length() const 67 | { 68 | return cpq.size(); 69 | } 70 | 71 | void clear() 72 | { 73 | std::unique_lock lock(mutex); 74 | while (!isEmpty()) 75 | { 76 | cpq.pop(); 77 | } 78 | lock.unlock(); 79 | cond.notify_all(); 80 | } 81 | }; 82 | 83 | #endif 84 | -------------------------------------------------------------------------------- /interleaving/Makefile: -------------------------------------------------------------------------------- 1 | test: raw_test.c 2 | gcc -o raw_test raw_test.c 3 | 4 | init: init.c 5 | gcc -o init -D_GNU_SOURCE init.c 6 | 7 | 8 | chunk: identify_chunk.c 9 | gcc -o identify_chunk -D_GNU_SOURCE identify_chunk.c 10 | 11 | interleave: identify_interleave.cpp 12 | g++ -std=c++11 -o identify_interleave -D_GNU_SOURCE identify_interleave.cpp -lpthread 13 | 14 | mapping: identify_mapping.cpp 15 | g++ -std=c++11 -o identify_mapping -D_GNU_SOURCE identify_mapping.cpp -lpthread 16 | 17 | split_request: split_request.cpp 18 | g++ -std=c++11 -o split_request -D_GNU_SOURCE split_request.cpp -lpthread 19 | 20 | shear_pattern: shear_identify_pattern.cpp 21 | g++ -std=c++11 -o shear_pattern -D_GNU_SOURCE shear_identify_pattern.cpp -lpthread 22 | 23 | random_overwrite: random_overwrite.cpp 24 | g++ -std=c++11 -o random_overwrite -D_GNU_SOURCE random_overwrite.cpp -lpthread 25 | 26 | multi_thread_aio: multi_thread_aio.cpp ConsumerProducerQueue.h 27 | g++ -std=c++11 -o multi_thread_aio -D_GNU_SOURCE multi_thread_aio.cpp ConsumerProducerQueue.h -lpthread -laio 28 | 29 | chunk_multi_thread_aio: multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h 30 | g++ -std=c++11 -o chunk_multi_thread_aio -D_GNU_SOURCE multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h -lpthread -laio 31 | 32 | pair_read: pair_read.cpp 33 | g++ -std=c++11 -o pair_read -D_GNU_SOURCE pair_read.cpp -lpthread 34 | 35 | clean: 36 | rm init raw_test identify_chunk identify_interleave identify_mapping 37 | -------------------------------------------------------------------------------- /interleaving/multi_thread_aio: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/sherlockwu/OptaneBench/90ef168cbd79cbfbc57122069b7cd79474bd325c/interleaving/multi_thread_aio -------------------------------------------------------------------------------- /interleaving/multi_thread_aio.cpp: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | #include 12 | #include 13 | #include 14 | #include 15 | #include 16 | #include 17 | #include 18 | #include "ConsumerProducerQueue.h" 19 | 20 | #define MAX_COUNT 65536 21 | 22 | std::atomic pointer(0); 23 | std::vector read_order; 24 | 25 | // Note: all pos in sector unless before passed in read/write 26 | int SECTOR_SIZE = 512; 27 | int STRIDE_SIZE = SECTOR_SIZE * 1; // chunk size actually 28 | int num_ios = 100000; 29 | int completed_ios = 0; 30 | 31 | int D = 0; 32 | int j = 0; 33 | int d = 0; 34 | int fd = 0; 35 | io_context_t ctx_; 36 | 37 | std::condition_variable cond; 38 | std::mutex mutex; 39 | int queue_depth = 0; 40 | int max_qd = 16; 41 | ConsumerProducerQueue job_queue; 42 | int pattern = 7; 43 | 44 | struct io_event events[MAX_COUNT]; 45 | struct timespec timeout; 46 | char * read_buf; 47 | 48 | #define handle_error_en(en, msg) \ 49 | do { errno = en; perror(msg); exit(EXIT_FAILURE); } while (0) 50 | 51 | void *eachThread(void *vargp) 52 | { 53 | // consumer 54 | int id = *(int*)vargp; 55 | 56 | // pin main thread to somewhere 57 | int s, j; 58 | cpu_set_t cpuset; 59 | pthread_t thread; 60 | 61 | thread = pthread_self(); 62 | 63 | /* Set affinity mask to include CPUs 0 to 4 */ 64 | CPU_ZERO(&cpuset); 65 | CPU_SET((id+1)%4, &cpuset); 66 | 67 | s = pthread_setaffinity_np(thread, sizeof(cpu_set_t), &cpuset); 68 | if (s != 0) 69 | handle_error_en(s, "pthread_setaffinity_np"); 70 | 71 | /* Check the actual affinity mask assigned to the thread */ 72 | s = pthread_getaffinity_np(thread, sizeof(cpu_set_t), &cpuset); 73 | if (s != 0) 74 | handle_error_en(s, "pthread_getaffinity_np"); 75 | 76 | printf("Set returned by pthread_getaffinity_np() contained:\n"); 77 | for (j = 0; j < CPU_SETSIZE; j++) 78 | if (CPU_ISSET(j, &cpuset)) 79 | printf(" CPU %d\n", j); 80 | 81 | 82 | long pos; // in unit of sector 83 | printf("Thread %d ready to run \n", id); 84 | 85 | int sz; 86 | 87 | for ( ; ; ) { 88 | // Consumer wait the queue depth control 89 | job_queue.consume(pos); 90 | if (pos == -1) { 91 | break; 92 | } 93 | //std::cout << "to issue IO " << pos/4096 << std::endl; 94 | struct iocb * p = (struct iocb *)malloc(sizeof(struct iocb)); 95 | //io_prep_pread(p, fd_[item.fd], item.buffer, item.length, item.offset); 96 | io_prep_pread(p, fd, read_buf, STRIDE_SIZE, pos); 97 | //p->data = (void *) item.io_status; 98 | p->data = (void *) read_buf; 99 | 100 | if (io_submit(ctx_, 1, &p) != 1) { 101 | io_destroy(ctx_); 102 | std::cout << "io submit error" << std::endl; 103 | exit(1); 104 | } 105 | } 106 | std::cout << "Thread " << id << " issued all IOs" << std::endl; 107 | return NULL; 108 | } 109 | 110 | 111 | void generate_read_trace(char type) { 112 | if (type == 'r') { 113 | std::cout << "To read randomly" << std::endl; 114 | srand(time(0)); 115 | for (int i = 0; i < num_ios; i++) 116 | read_order.push_back((long)((rand() %(80*1024*1024*2/d)))*STRIDE_SIZE); 117 | } else if (type == 's') { 118 | std::cout << "To read sequentially with offset " << D << std::endl; 119 | for (int i = 0; i < num_ios; i++) { 120 | // who is channel 1? 121 | // i*7 is in: (0+D*i)%7 122 | //std::cout << i*7 << " in " << (D*i)%7 << " first channel: " << i*7 + (7-((D*i)%7-0)) %7 << std::endl; 123 | //read_order.push_back((long)(i*7 + (7-((D*i)%7-0)) %7)*STRIDE_SIZE); 124 | read_order.push_back((long)(i*STRIDE_SIZE)); 125 | } 126 | //exit(1); 127 | //for (int i = 0; i < num_ios; i++) 128 | // read_order.push_back((long)(i)*STRIDE_SIZE); 129 | } else if (type == 'j') { 130 | std::cout << "To read stridely with stride: " << D << std::endl; 131 | for (int i = 0; i < num_ios; i++) 132 | read_order.push_back(((long)(i+0)*(STRIDE_SIZE + D*SECTOR_SIZE))%((long)800*1024*1024*1024)); 133 | } else if (type == 'z') { 134 | std::cout << "To read zigzag" << std::endl; 135 | for (int i = 0; i < num_ios; i++) 136 | read_order.push_back((long)(i*7 + (i%2)*6)*STRIDE_SIZE); 137 | } else if (type == 'p') { // shear pair 138 | std::cout << "To read according to shear pair(randomly choose pattern, then, issue two parallel reads within each pattern at the same offsets)" << std::endl; 139 | srand(time(0)); 140 | std::cout << "Pattern size: " << pattern << std::endl; 141 | for (int i = 0; i < num_ios-1; i += 2) { 142 | // segment number 143 | long segment = rand() % (800*1024*1024*2/(d*pattern)); 144 | read_order.push_back(segment*(pattern*STRIDE_SIZE)); 145 | read_order.push_back(segment*(pattern*STRIDE_SIZE) + (D+d)*SECTOR_SIZE); 146 | } 147 | } else if (type == 'e') { // split sequential 8 sectors into 2*4 sectors etc 148 | std::cout << "To read stridely with stride: " << D << " and split 8 sectors "<< std::endl; 149 | int small = 8/ (STRIDE_SIZE/SECTOR_SIZE); 150 | num_ios *= small; 151 | for (int i = 0; i < num_ios/small; i++) { 152 | for (int j = 0; j < small; j++) 153 | read_order.push_back((long)(i)*(8*SECTOR_SIZE + D*SECTOR_SIZE) + j*STRIDE_SIZE); 154 | } 155 | } else if (type == '1') { // split 8 sectors into 1*QD sectors 156 | std::cout << "To read stridely with stride: " << D << " and choose " << max_qd << " out of each 8 sectors "<< std::endl; 157 | int small = max_qd; 158 | num_ios *= small; 159 | for (int i = 0; i < num_ios/small; i++) { 160 | for (int j = 0; j < small; j++) 161 | read_order.push_back((long)(i)*(8*SECTOR_SIZE + D*SECTOR_SIZE) + j*STRIDE_SIZE); 162 | } 163 | } else if (type == '4') { // split 8 sectors into 1*QD sectors 164 | int small = 4; 165 | num_ios *= small; 166 | for (int i = 0; i < num_ios/small; i++) { 167 | for (int j = 0; j < small; j++) 168 | read_order.push_back((long)(i)*(8*SECTOR_SIZE + D*SECTOR_SIZE) + j*STRIDE_SIZE); 169 | } 170 | } else { 171 | std::cout << "Wrong choice of workloads, should be seq/random" << std::endl; 172 | exit(1); 173 | } 174 | 175 | return; 176 | } 177 | 178 | int main(int argc, char* * argv) { 179 | // pin main thread to somewhere 180 | int s, j; 181 | cpu_set_t cpuset; 182 | pthread_t thread; 183 | 184 | thread = pthread_self(); 185 | 186 | /* Set affinity mask to include CPUs 0 to 3 */ 187 | CPU_ZERO(&cpuset); 188 | CPU_SET(0, &cpuset); 189 | 190 | s = pthread_setaffinity_np(thread, sizeof(cpu_set_t), &cpuset); 191 | if (s != 0) 192 | handle_error_en(s, "pthread_setaffinity_np"); 193 | 194 | /* Check the actual affinity mask assigned to the thread */ 195 | s = pthread_getaffinity_np(thread, sizeof(cpu_set_t), &cpuset); 196 | if (s != 0) 197 | handle_error_en(s, "pthread_getaffinity_np"); 198 | 199 | printf("Set returned by pthread_getaffinity_np() contained:\n"); 200 | for (j = 0; j < CPU_SETSIZE; j++) 201 | if (CPU_ISSET(j, &cpuset)) 202 | printf(" CPU %d\n", j); 203 | 204 | 205 | printf("===== Multi-thread libaio to Specified Device =====\n"); 206 | // identify chunk size 207 | if (argc < 7) { 208 | printf("Wrong parameters: multi_thread_aio dev_name D(offset to jump, in unit of sector) j(num_parallel_jobs to submit IO) d(stride/io_size in sector) read_type(random/seq/jump) queue_depth\n"); 209 | return 1; 210 | } 211 | D = atoi(argv[2]); //offset between two reads, in unit of sector 212 | j = atoi(argv[3]); 213 | d = atoi(argv[4]); // request size 214 | STRIDE_SIZE = SECTOR_SIZE * d; 215 | max_qd = atoi(argv[6]); 216 | //max_qd = 16; 217 | pattern = atoi(argv[6])/16; 218 | job_queue.set_max(max_qd); 219 | 220 | printf("To run with:\n jump: %d, num_jobs: %d, stride: %d, device_name: %s, write_order: %s, max_qd: %d\n", D, j, d, argv[1], argv[5], max_qd); 221 | 222 | // open raw block device 223 | fd = open(argv[1], O_RDONLY | O_DIRECT); // O_DIRECT 224 | if (fd < 0) { 225 | printf("Raw Device Open failed\n"); 226 | return 1; 227 | } 228 | // context to do async IO 229 | memset(&ctx_, 0, sizeof(ctx_)); 230 | if (io_setup(MAX_COUNT, &ctx_) != 0) { 231 | std::cout << "io_context_t set failed" << std::endl; 232 | exit(1); 233 | } 234 | // generate read trace 235 | generate_read_trace(argv[5][0]); 236 | 237 | //parallel reads 238 | 239 | // start the number of threads 240 | pthread_t * thread_pool = (pthread_t *)malloc(sizeof(pthread_t) * j); 241 | 242 | 243 | for (int i = 0; i < j; i++) { 244 | int *arg = (int *)malloc(sizeof(*arg)); 245 | *arg = i; 246 | pthread_create(&thread_pool[i], NULL, eachThread, (void *)arg); 247 | } 248 | 249 | 250 | read_buf = (char *) malloc(sizeof(char) * (STRIDE_SIZE + SECTOR_SIZE)); 251 | int ret = posix_memalign((void **)&read_buf, SECTOR_SIZE, STRIDE_SIZE + SECTOR_SIZE); 252 | sleep(1); 253 | 254 | //as a producer 255 | struct timeval start, end; 256 | gettimeofday(&start, NULL); 257 | int onflight_io = 0; 258 | 259 | timeout.tv_sec = 0; 260 | timeout.tv_nsec = 100; 261 | for (int i = 0; i < num_ios;) { 262 | // just add 263 | int ret = io_getevents(ctx_, 0, MAX_COUNT, events, &timeout); 264 | if (ret < 0) { 265 | std::cout << "Getevents Error" << std::endl; 266 | exit(1); 267 | } 268 | if (ret > 0) { 269 | onflight_io -= ret; 270 | //std::cout << "=== Get Events("<< ret << "): " << ret << std::endl; 271 | //for (int j = 0; j < ret; j++) { 272 | // std::cout << "=== Complete " << "::" << events[j].obj->aio_fildes << std::endl; 273 | // free((struct iocb *)events[j].obj); 274 | //} 275 | } 276 | 277 | //if (onflight_io >= max_qd) 278 | //continue; 279 | 280 | //std::cout << "Add " << i << ": " << read_order[i] << std::endl; 281 | //if (argv[5][0] == 'j') { 282 | if (onflight_io >= max_qd) 283 | //if (onflight_io/2 >= max_qd) 284 | continue; 285 | 286 | while (onflight_io < max_qd && i < num_ios) { 287 | job_queue.add(read_order[i]); 288 | onflight_io++; 289 | i++; 290 | } 291 | //job_queue.add(read_order[i]); 292 | //job_queue.add(read_order[i] + SECTOR_SIZE); 293 | //onflight_io+=2; 294 | //onflight_io++; 295 | /*} else { 296 | if (onflight_io >= max_qd) 297 | continue; 298 | job_queue.add(read_order[i]); 299 | onflight_io++; 300 | i++; 301 | }*/ 302 | //i++; 303 | } 304 | 305 | //wait until all IO finished 306 | while (onflight_io) { 307 | int ret = io_getevents(ctx_, 0, MAX_COUNT, events, &timeout); 308 | if (ret < 0) { 309 | std::cout << "Getevents Error" << std::endl; 310 | exit(1); 311 | } 312 | if (ret > 0) { 313 | onflight_io -= ret; 314 | //std::cout << "=== Get Events("<< ret << "): " << ret << std::endl; 315 | //for (int j = 0; j < ret; j++) { 316 | // std::cout << "=== Complete " << "::" << events[j].obj->aio_fildes << std::endl; 317 | //} 318 | } 319 | } 320 | 321 | 322 | 323 | gettimeofday(&end, NULL); 324 | long time_us = ((end.tv_sec * 1000000 + end.tv_usec) 325 | - (start.tv_sec * 1000000 + start.tv_usec)); 326 | 327 | 328 | // Signal to exit 329 | for (int i = 0; i < j; i++) { 330 | job_queue.add(-1); 331 | } 332 | for (int i = 0; i < j; i++) { 333 | pthread_join(thread_pool[i], NULL); 334 | } 335 | std::cout << "All IO finished " << STRIDE_SIZE << std::endl; 336 | printf("Time taken: %ld us, Bandwidth: %f MB/s \n", time_us, (float)((long)STRIDE_SIZE*num_ios)/1024/1024*1000000/time_us); 337 | 338 | free(read_buf); 339 | close(fd); 340 | return 0; 341 | } 342 | -------------------------------------------------------------------------------- /mapping/Makefile: -------------------------------------------------------------------------------- 1 | test: raw_test.c 2 | gcc -o raw_test raw_test.c 3 | 4 | init: init.c 5 | gcc -o init -D_GNU_SOURCE init.c 6 | 7 | 8 | chunk: identify_chunk.c 9 | gcc -o identify_chunk -D_GNU_SOURCE identify_chunk.c 10 | 11 | interleave: identify_interleave.cpp 12 | g++ -std=c++11 -o identify_interleave -D_GNU_SOURCE identify_interleave.cpp -lpthread 13 | 14 | mapping: identify_mapping.cpp 15 | g++ -std=c++11 -o identify_mapping -D_GNU_SOURCE identify_mapping.cpp -lpthread 16 | 17 | split_request: split_request.cpp 18 | g++ -std=c++11 -o split_request -D_GNU_SOURCE split_request.cpp -lpthread 19 | 20 | shear_pattern: shear_identify_pattern.cpp 21 | g++ -std=c++11 -o shear_pattern -D_GNU_SOURCE shear_identify_pattern.cpp -lpthread 22 | 23 | random_overwrite: random_overwrite.cpp 24 | g++ -std=c++11 -o random_overwrite -D_GNU_SOURCE random_overwrite.cpp -lpthread 25 | 26 | multi_thread_aio: multi_thread_aio.cpp ConsumerProducerQueue.h 27 | g++ -std=c++11 -o multi_thread_aio -D_GNU_SOURCE multi_thread_aio.cpp ConsumerProducerQueue.h -lpthread -laio 28 | 29 | chunk_multi_thread_aio: multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h 30 | g++ -std=c++11 -o chunk_multi_thread_aio -D_GNU_SOURCE multi_thread_aio_for_chunk.cpp ConsumerProducerQueue.h -lpthread -laio 31 | 32 | pair_read: pair_read.cpp 33 | g++ -std=c++11 -o pair_read -D_GNU_SOURCE pair_read.cpp -lpthread 34 | 35 | clean: 36 | rm init raw_test identify_chunk identify_interleave identify_mapping 37 | -------------------------------------------------------------------------------- /mapping/identify_mapping.cpp: -------------------------------------------------------------------------------- 1 | #define _GNU_SOURCE 2 | 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | #include 12 | #include 13 | #include 14 | #include 15 | #include 16 | #include 17 | 18 | std::atomic pointer(0); 19 | std::vector write_order; 20 | 21 | // Note: all pos in sector unless before passed in read/write 22 | int SECTOR_SIZE = 512; 23 | int STRIDE_SIZE = SECTOR_SIZE * 16; // chunk size actually 24 | int num_ios = 10000; 25 | 26 | int jobs = 0; 27 | int j = 0; 28 | int d = 0; 29 | int fd = 0; 30 | void *eachThread(void *vargp) 31 | { 32 | int id = *(int*)vargp; 33 | long pos; // in unit of sector 34 | printf("Thread %d ready to run \n", id); 35 | char * read_buf = (char *) malloc(sizeof(char) * (STRIDE_SIZE + SECTOR_SIZE)); 36 | int ret = posix_memalign((void **)&read_buf, SECTOR_SIZE, STRIDE_SIZE + SECTOR_SIZE); 37 | 38 | int sz; 39 | while (jobs <= 0) ; 40 | 41 | // do IO 42 | 43 | for ( ; ; ) { 44 | pos = pointer.fetch_add(1); 45 | if (pos > num_ios) 46 | break; 47 | //printf("Position: %d \n", pos * d); 48 | //sz = pread(fd, read_buf, 8 * SECTOR_SIZE, (pos + i * big_stride)* SECTOR_SIZE); 49 | //sz = pread(fd, read_buf, 8 * SECTOR_SIZE, pointer * d * SECTOR_SIZE); 50 | //assert(sz == 8*SECTOR_SIZE); 51 | //sz = pread(fd, read_buf, STRIDE_SIZE, pos * d * STRIDE_SIZE); 52 | //sz = pread(fd, read_buf, STRIDE_SIZE, (pos/j) * 16 * STRIDE_SIZE + (pos%j) * d * STRIDE_SIZE); 53 | //sz = pread(fd, read_buf, STRIDE_SIZE, (pos%j) * d * STRIDE_SIZE); 54 | //sz = pread(fd, read_buf, STRIDE_SIZE, (pos%2) * d * STRIDE_SIZE); 55 | //sz = pread(fd, read_buf, STRIDE_SIZE, (pos%2) * d * STRIDE_SIZE); 56 | // according to write order 57 | //std::cout << pos << ":" << write_order[(pos%2) * d] << std::endl; 58 | //sz = pread(fd, read_buf, STRIDE_SIZE, write_order[(pos%2) * d]); 59 | sz = pread(fd, read_buf, STRIDE_SIZE, write_order[pos * (d+1)]); 60 | //printf("Done %ld %d\n", pointer*d*STRIDE_SIZE, sz); 61 | assert(sz == STRIDE_SIZE); 62 | } 63 | free(read_buf); 64 | return NULL; 65 | } 66 | 67 | 68 | 69 | int main(int argc, char* * argv) { 70 | printf("===== Identify Mapping of Specified Device(read according to writes order) =====\n"); 71 | // identify chunk size 72 | if (argc < 6) { 73 | printf("Wrong parameters: identify_mapping dev_name D(Estimated Max interleave degree) j(num_parallel_jobs) d(stride in chunk(16 sectors)) write_order(write order file name)\n"); 74 | return 1; 75 | } 76 | int D = atoi(argv[2]); 77 | j = atoi(argv[3]); 78 | d = atoi(argv[4]); 79 | 80 | printf("To run with:\n Max_Degree: %d, num_jobs: %d, stride: %d, device_name: %s, write_order: %s\n", D, j, d, argv[1], argv[5]); 81 | 82 | // open raw block device 83 | fd = open(argv[1], O_RDONLY | O_DIRECT); // O_DIRECT 84 | if (fd < 0) { 85 | printf("Raw Device Open failed\n"); 86 | return 1; 87 | } 88 | 89 | //parallel reads 90 | 91 | // start the number of threads 92 | pthread_t * thread_pool = (pthread_t *)malloc(sizeof(pthread_t) * j); 93 | 94 | for (int i = 0; i < j; i++) { 95 | int *arg = (int *)malloc(sizeof(*arg)); 96 | *arg = i; 97 | pthread_create(&thread_pool[i], NULL, eachThread, (void *)arg); 98 | } 99 | 100 | // prepare the read trace (according to pointer) 101 | std::ifstream write_trace(argv[5]); 102 | long cur_write; 103 | while(write_trace >> cur_write) { 104 | write_order.push_back(cur_write); 105 | } 106 | 107 | sleep(2); 108 | 109 | struct timeval start, end; 110 | gettimeofday(&start, NULL); 111 | jobs = j; // hint to start doing IOs 112 | for (int i = 0; i < j; i++) { 113 | pthread_join(thread_pool[i], NULL); 114 | } 115 | gettimeofday(&end, NULL); 116 | long time_us = ((end.tv_sec * 1000000 + end.tv_usec) 117 | - (start.tv_sec * 1000000 + start.tv_usec)); 118 | //printf("Time taken: %ld us, Bandwidth: %f MB/s \n", time_us, (float)(8*512*100000)/1024/1024*1000000/time_us); 119 | printf("Time taken: %ld us, Bandwidth: %f MB/s \n", time_us, (float)(STRIDE_SIZE*num_ios)/1024/1024*1000000/time_us); 120 | 121 | close(fd); 122 | return 0; 123 | } 124 | -------------------------------------------------------------------------------- /mapping/random_overwrite.cpp: -------------------------------------------------------------------------------- 1 | #define _GNU_SOURCE 2 | 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | #include 12 | #include 13 | #include 14 | // Note: all pos in sector unless before passed in read/write 15 | int SECTOR_SIZE = 512; 16 | int S = 0; 17 | int n = 0; 18 | 19 | 20 | int main(int argc, char* * argv) { 21 | // identify chunk size 22 | if (argc < 4) { 23 | printf("Wrong parameters: random_overwrite dev_name S(overwritten zone in GB, start from 0) n(req_size in sector) \n"); 24 | return 1; 25 | } 26 | S = atoi(argv[2]); 27 | n = atoi(argv[3]); 28 | 29 | char * write_buf = (char *) malloc(sizeof(char) * (n * SECTOR_SIZE)); 30 | int ret = posix_memalign((void **)&write_buf, SECTOR_SIZE, n*SECTOR_SIZE); 31 | 32 | //printf("To run with:\n zone_size: %d GB, request_size: %d, device_name: %s\n", S, n, argv[1]); 33 | 34 | // open raw block device 35 | int fd = open(argv[1], O_WRONLY | O_DIRECT); 36 | if (fd < 0) { 37 | printf("Raw Device Open failed\n"); 38 | return 1; 39 | } 40 | 41 | //generate_random_overwrite_order(); 42 | std::vector pos_vector; 43 | for (long i = 0; i < (long)(S)*1024*1024*1024; i+=n*SECTOR_SIZE) pos_vector.push_back(i); 44 | std::random_shuffle (pos_vector.begin(), pos_vector.end() ); 45 | std::cout << pos_vector.size() << std::endl; 46 | for (auto it = pos_vector.begin(); it != pos_vector.end(); it++) { 47 | std::cout << *it << std::endl; 48 | int sz = pwrite(fd, write_buf, n * SECTOR_SIZE, *it); 49 | assert(sz == n*SECTOR_SIZE); 50 | } 51 | 52 | 53 | //for (int i = 0; i < S; i+=n) { 54 | // long pos = rand_pos(M); 55 | // int sz = pwrite(fd, write_buf, n * SECTOR_SIZE); 56 | // assert(sz == n*SECTOR_SIZE); 57 | //} 58 | 59 | free(write_buf); 60 | close(fd); 61 | return 0; 62 | } 63 | -------------------------------------------------------------------------------- /multi_thread_aio.cpp: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | #include 12 | #include 13 | #include 14 | #include 15 | #include 16 | #include 17 | #include 18 | #include "ConsumerProducerQueue.h" 19 | 20 | #define MAX_COUNT 65536 21 | 22 | std::atomic pointer(0); 23 | std::vector read_order; 24 | 25 | // Note: all pos in sector unless before passed in read/write 26 | int SECTOR_SIZE = 512; 27 | int STRIDE_SIZE = SECTOR_SIZE * 1; // chunk size actually 28 | int num_ios = 100000; 29 | int completed_ios = 0; 30 | 31 | int D = 0; 32 | int j = 0; 33 | int d = 0; 34 | int fd = 0; 35 | io_context_t ctx_; 36 | 37 | std::condition_variable cond; 38 | std::mutex mutex; 39 | int queue_depth = 0; 40 | int max_qd = 16; 41 | ConsumerProducerQueue job_queue; 42 | int pattern = 7; 43 | 44 | struct io_event events[MAX_COUNT]; 45 | struct timespec timeout; 46 | char * read_buf; 47 | 48 | #define handle_error_en(en, msg) \ 49 | do { errno = en; perror(msg); exit(EXIT_FAILURE); } while (0) 50 | 51 | void *eachThread(void *vargp) 52 | { 53 | // consumer 54 | int id = *(int*)vargp; 55 | 56 | // pin main thread to somewhere 57 | int s, j; 58 | cpu_set_t cpuset; 59 | pthread_t thread; 60 | 61 | thread = pthread_self(); 62 | 63 | /* Set affinity mask to include CPUs 0 to 4 */ 64 | CPU_ZERO(&cpuset); 65 | CPU_SET((id+1)%4, &cpuset); 66 | 67 | s = pthread_setaffinity_np(thread, sizeof(cpu_set_t), &cpuset); 68 | if (s != 0) 69 | handle_error_en(s, "pthread_setaffinity_np"); 70 | 71 | /* Check the actual affinity mask assigned to the thread */ 72 | s = pthread_getaffinity_np(thread, sizeof(cpu_set_t), &cpuset); 73 | if (s != 0) 74 | handle_error_en(s, "pthread_getaffinity_np"); 75 | 76 | printf("Set returned by pthread_getaffinity_np() contained:\n"); 77 | for (j = 0; j < CPU_SETSIZE; j++) 78 | if (CPU_ISSET(j, &cpuset)) 79 | printf(" CPU %d\n", j); 80 | 81 | 82 | long pos; // in unit of sector 83 | printf("Thread %d ready to run \n", id); 84 | 85 | int sz; 86 | 87 | for ( ; ; ) { 88 | // Consumer wait the queue depth control 89 | job_queue.consume(pos); 90 | if (pos == -1) { 91 | break; 92 | } 93 | //std::cout << "to issue IO " << pos/4096 << std::endl; 94 | struct iocb * p = (struct iocb *)malloc(sizeof(struct iocb)); 95 | //io_prep_pread(p, fd_[item.fd], item.buffer, item.length, item.offset); 96 | io_prep_pread(p, fd, read_buf, STRIDE_SIZE, pos); 97 | //p->data = (void *) item.io_status; 98 | p->data = (void *) read_buf; 99 | 100 | if (io_submit(ctx_, 1, &p) != 1) { 101 | io_destroy(ctx_); 102 | std::cout << "io submit error" << std::endl; 103 | exit(1); 104 | } 105 | } 106 | std::cout << "Thread " << id << " issued all IOs" << std::endl; 107 | return NULL; 108 | } 109 | 110 | 111 | void generate_read_trace(char type) { 112 | if (type == 'r') { 113 | std::cout << "To read randomly" << std::endl; 114 | srand(time(0)); 115 | for (int i = 0; i < num_ios; i++) 116 | read_order.push_back((long)((rand() %(80*1024*1024*2/d)))*STRIDE_SIZE); 117 | } else if (type == 's') { 118 | std::cout << "To read sequentially with offset " << D << std::endl; 119 | for (int i = 0; i < num_ios; i++) { 120 | // who is channel 1? 121 | // i*7 is in: (0+D*i)%7 122 | //std::cout << i*7 << " in " << (D*i)%7 << " first channel: " << i*7 + (7-((D*i)%7-0)) %7 << std::endl; 123 | //read_order.push_back((long)(i*7 + (7-((D*i)%7-0)) %7)*STRIDE_SIZE); 124 | read_order.push_back((long)(i*STRIDE_SIZE)); 125 | } 126 | //exit(1); 127 | //for (int i = 0; i < num_ios; i++) 128 | // read_order.push_back((long)(i)*STRIDE_SIZE); 129 | } else if (type == 'j') { 130 | std::cout << "To read stridely with stride: " << D << std::endl; 131 | for (int i = 0; i < num_ios; i++) 132 | read_order.push_back(((long)(i+0)*(STRIDE_SIZE + D*SECTOR_SIZE))%((long)800*1024*1024*1024)); 133 | } else if (type == 'z') { 134 | std::cout << "To read zigzag" << std::endl; 135 | for (int i = 0; i < num_ios; i++) 136 | read_order.push_back((long)(i*7 + (i%2)*6)*STRIDE_SIZE); 137 | } else if (type == 'p') { // shear pair 138 | std::cout << "To read according to shear pair(randomly choose pattern, then, issue two parallel reads within each pattern at the same offsets)" << std::endl; 139 | srand(time(0)); 140 | std::cout << "Pattern size: " << pattern << std::endl; 141 | for (int i = 0; i < num_ios-1; i += 2) { 142 | // segment number 143 | long segment = rand() % (800*1024*1024*2/(d*pattern)); 144 | read_order.push_back(segment*(pattern*STRIDE_SIZE)); 145 | read_order.push_back(segment*(pattern*STRIDE_SIZE) + (D+d)*SECTOR_SIZE); 146 | } 147 | } else if (type == 'e') { // split sequential 8 sectors into 2*4 sectors etc 148 | std::cout << "To read stridely with stride: " << D << " and split 8 sectors "<< std::endl; 149 | int small = 8/ (STRIDE_SIZE/SECTOR_SIZE); 150 | num_ios *= small; 151 | for (int i = 0; i < num_ios/small; i++) { 152 | for (int j = 0; j < small; j++) 153 | read_order.push_back((long)(i)*(8*SECTOR_SIZE + D*SECTOR_SIZE) + j*STRIDE_SIZE); 154 | } 155 | } else if (type == '1') { // split 8 sectors into 1*QD sectors 156 | std::cout << "To read stridely with stride: " << D << " and choose " << max_qd << " out of each 8 sectors "<< std::endl; 157 | int small = max_qd; 158 | num_ios *= small; 159 | for (int i = 0; i < num_ios/small; i++) { 160 | for (int j = 0; j < small; j++) 161 | read_order.push_back((long)(i)*(8*SECTOR_SIZE + D*SECTOR_SIZE) + j*STRIDE_SIZE); 162 | } 163 | } else if (type == '4') { // split 8 sectors into 1*QD sectors 164 | int small = 4; 165 | num_ios *= small; 166 | for (int i = 0; i < num_ios/small; i++) { 167 | for (int j = 0; j < small; j++) 168 | read_order.push_back((long)(i)*(8*SECTOR_SIZE + D*SECTOR_SIZE) + j*STRIDE_SIZE); 169 | } 170 | } else { 171 | std::cout << "Wrong choice of workloads, should be seq/random" << std::endl; 172 | exit(1); 173 | } 174 | 175 | return; 176 | } 177 | 178 | int main(int argc, char* * argv) { 179 | // pin main thread to somewhere 180 | int s, j; 181 | cpu_set_t cpuset; 182 | pthread_t thread; 183 | 184 | thread = pthread_self(); 185 | 186 | /* Set affinity mask to include CPUs 0 to 3 */ 187 | CPU_ZERO(&cpuset); 188 | CPU_SET(0, &cpuset); 189 | 190 | s = pthread_setaffinity_np(thread, sizeof(cpu_set_t), &cpuset); 191 | if (s != 0) 192 | handle_error_en(s, "pthread_setaffinity_np"); 193 | 194 | /* Check the actual affinity mask assigned to the thread */ 195 | s = pthread_getaffinity_np(thread, sizeof(cpu_set_t), &cpuset); 196 | if (s != 0) 197 | handle_error_en(s, "pthread_getaffinity_np"); 198 | 199 | printf("Set returned by pthread_getaffinity_np() contained:\n"); 200 | for (j = 0; j < CPU_SETSIZE; j++) 201 | if (CPU_ISSET(j, &cpuset)) 202 | printf(" CPU %d\n", j); 203 | 204 | 205 | printf("===== Multi-thread libaio to Specified Device =====\n"); 206 | // identify chunk size 207 | if (argc < 7) { 208 | printf("Wrong parameters: multi_thread_aio dev_name D(offset to jump, in unit of sector) j(num_parallel_jobs to submit IO) d(stride/io_size in sector) read_type(random/seq/jump) queue_depth\n"); 209 | return 1; 210 | } 211 | D = atoi(argv[2]); //offset between two reads, in unit of sector 212 | j = atoi(argv[3]); 213 | d = atoi(argv[4]); // request size 214 | STRIDE_SIZE = SECTOR_SIZE * d; 215 | max_qd = atoi(argv[6]); 216 | //max_qd = 16; 217 | pattern = atoi(argv[6])/16; 218 | job_queue.set_max(max_qd); 219 | 220 | printf("To run with:\n jump: %d, num_jobs: %d, stride: %d, device_name: %s, write_order: %s, max_qd: %d\n", D, j, d, argv[1], argv[5], max_qd); 221 | 222 | // open raw block device 223 | fd = open(argv[1], O_RDONLY | O_DIRECT); // O_DIRECT 224 | if (fd < 0) { 225 | printf("Raw Device Open failed\n"); 226 | return 1; 227 | } 228 | // context to do async IO 229 | memset(&ctx_, 0, sizeof(ctx_)); 230 | if (io_setup(MAX_COUNT, &ctx_) != 0) { 231 | std::cout << "io_context_t set failed" << std::endl; 232 | exit(1); 233 | } 234 | // generate read trace 235 | generate_read_trace(argv[5][0]); 236 | 237 | //parallel reads 238 | 239 | // start the number of threads 240 | pthread_t * thread_pool = (pthread_t *)malloc(sizeof(pthread_t) * j); 241 | 242 | 243 | for (int i = 0; i < j; i++) { 244 | int *arg = (int *)malloc(sizeof(*arg)); 245 | *arg = i; 246 | pthread_create(&thread_pool[i], NULL, eachThread, (void *)arg); 247 | } 248 | 249 | 250 | read_buf = (char *) malloc(sizeof(char) * (STRIDE_SIZE + SECTOR_SIZE)); 251 | int ret = posix_memalign((void **)&read_buf, SECTOR_SIZE, STRIDE_SIZE + SECTOR_SIZE); 252 | sleep(1); 253 | 254 | //as a producer 255 | struct timeval start, end; 256 | gettimeofday(&start, NULL); 257 | int onflight_io = 0; 258 | 259 | timeout.tv_sec = 0; 260 | timeout.tv_nsec = 100; 261 | for (int i = 0; i < num_ios;) { 262 | // just add 263 | int ret = io_getevents(ctx_, 0, MAX_COUNT, events, &timeout); 264 | if (ret < 0) { 265 | std::cout << "Getevents Error" << std::endl; 266 | exit(1); 267 | } 268 | if (ret > 0) { 269 | onflight_io -= ret; 270 | //std::cout << "=== Get Events("<< ret << "): " << ret << std::endl; 271 | //for (int j = 0; j < ret; j++) { 272 | // std::cout << "=== Complete " << "::" << events[j].obj->aio_fildes << std::endl; 273 | // free((struct iocb *)events[j].obj); 274 | //} 275 | } 276 | 277 | //if (onflight_io >= max_qd) 278 | //continue; 279 | 280 | //std::cout << "Add " << i << ": " << read_order[i] << std::endl; 281 | //if (argv[5][0] == 'j') { 282 | if (onflight_io >= max_qd) 283 | //if (onflight_io/2 >= max_qd) 284 | continue; 285 | 286 | while (onflight_io < max_qd && i < num_ios) { 287 | job_queue.add(read_order[i]); 288 | onflight_io++; 289 | i++; 290 | } 291 | //job_queue.add(read_order[i]); 292 | //job_queue.add(read_order[i] + SECTOR_SIZE); 293 | //onflight_io+=2; 294 | //onflight_io++; 295 | /*} else { 296 | if (onflight_io >= max_qd) 297 | continue; 298 | job_queue.add(read_order[i]); 299 | onflight_io++; 300 | i++; 301 | }*/ 302 | //i++; 303 | } 304 | 305 | //wait until all IO finished 306 | while (onflight_io) { 307 | int ret = io_getevents(ctx_, 0, MAX_COUNT, events, &timeout); 308 | if (ret < 0) { 309 | std::cout << "Getevents Error" << std::endl; 310 | exit(1); 311 | } 312 | if (ret > 0) { 313 | onflight_io -= ret; 314 | //std::cout << "=== Get Events("<< ret << "): " << ret << std::endl; 315 | //for (int j = 0; j < ret; j++) { 316 | // std::cout << "=== Complete " << "::" << events[j].obj->aio_fildes << std::endl; 317 | //} 318 | } 319 | } 320 | 321 | 322 | 323 | gettimeofday(&end, NULL); 324 | long time_us = ((end.tv_sec * 1000000 + end.tv_usec) 325 | - (start.tv_sec * 1000000 + start.tv_usec)); 326 | 327 | 328 | // Signal to exit 329 | for (int i = 0; i < j; i++) { 330 | job_queue.add(-1); 331 | } 332 | for (int i = 0; i < j; i++) { 333 | pthread_join(thread_pool[i], NULL); 334 | } 335 | std::cout << "All IO finished " << STRIDE_SIZE << std::endl; 336 | printf("Time taken: %ld us, Bandwidth: %f MB/s \n", time_us, (float)((long)STRIDE_SIZE*num_ios)/1024/1024*1000000/time_us); 337 | 338 | free(read_buf); 339 | close(fd); 340 | return 0; 341 | } 342 | -------------------------------------------------------------------------------- /multi_thread_aio_for_pair.cpp: -------------------------------------------------------------------------------- 1 | #define _GNU_SOURCE 2 | 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | #include 12 | #include 13 | #include 14 | #include 15 | #include 16 | #include 17 | #include 18 | #include 19 | #include 20 | #include "ConsumerProducerQueue.h" 21 | 22 | #define MAX_COUNT 65536 23 | 24 | std::atomic pointer(0); 25 | std::vector read_order; 26 | 27 | // Note: all pos in sector unless before passed in read/write 28 | int SECTOR_SIZE = 512; 29 | int STRIDE_SIZE = SECTOR_SIZE * 1; // chunk size actually 30 | int num_ios = 100000; 31 | int completed_ios = 0; 32 | 33 | int D = 0; 34 | int j = 0; 35 | int d = 0; 36 | int fd = 0; 37 | io_context_t ctx_; 38 | int TRIES = 5; // for shear 39 | long overall_us = 0; 40 | 41 | std::condition_variable cond; 42 | std::mutex mutex; 43 | int queue_depth = 0; 44 | int max_qd = 16; 45 | ConsumerProducerQueue job_queue; 46 | int pattern = 7; 47 | 48 | struct io_event events[MAX_COUNT]; 49 | struct timespec timeout; 50 | char * read_buf; 51 | 52 | void *eachThread(void *vargp) 53 | { 54 | // consumer 55 | int id = *(int*)vargp; 56 | long pos; // in unit of sector 57 | printf("Thread %d ready to run \n", id); 58 | 59 | int sz; 60 | 61 | for ( ; ; ) { 62 | // Consumer wait the queue depth control 63 | job_queue.consume(pos); 64 | if (pos == -1) { 65 | break; 66 | } 67 | //std::cout << "to issue IO " << pos/4096 << std::endl; 68 | struct iocb * p = (struct iocb *)malloc(sizeof(struct iocb)); 69 | //io_prep_pread(p, fd_[item.fd], item.buffer, item.length, item.offset); 70 | io_prep_pread(p, fd, read_buf, STRIDE_SIZE, pos); 71 | //p->data = (void *) item.io_status; 72 | p->data = (void *) read_buf; 73 | 74 | if (io_submit(ctx_, 1, &p) != 1) { 75 | io_destroy(ctx_); 76 | std::cout << "io submit error" << std::endl; 77 | exit(1); 78 | } 79 | } 80 | std::cout << "Thread " << id << " issued all IOs" << std::endl; 81 | return NULL; 82 | } 83 | 84 | 85 | void generate_read_trace(char type) { 86 | if (type == 's') { 87 | std::cout << "running shear: qd: " << max_qd << " segment size(in sector): " << D*d << std::endl; 88 | srand(time(0)); 89 | for (int i = 0; i < num_ios; i++) // generate random segments 90 | read_order.push_back((long)((rand() %(800*1024*1024*2/(D*d))))*D*STRIDE_SIZE); 91 | return; 92 | } else if (type == 'r') { 93 | std::cout << "To read randomly" << std::endl; 94 | srand(time(0)); 95 | for (int i = 0; i < num_ios; i++) 96 | read_order.push_back((long)((rand() %(800*1024*1024/4)))*4096); 97 | } else { 98 | std::cout << "Wrong choice of workloads, should be seq/random" << std::endl; 99 | exit(1); 100 | } 101 | 102 | return; 103 | } 104 | 105 | int main(int argc, char* * argv) { 106 | printf("===== Multi-thread libaio to Specified Device =====\n"); 107 | // identify chunk size 108 | if (argc < 7) { 109 | printf("Wrong parameters: multi_thread_aio dev_name D(offset to jump, in unit of sector) j(num_parallel_jobs to submit IO) d(stride/io_size in sector) read_type(random/seq/jump) queue_depth\n"); 110 | return 1; 111 | } 112 | D = atoi(argv[2]); //offset between two reads, in unit of sector 113 | j = atoi(argv[3]); 114 | d = atoi(argv[4]); // request size 115 | STRIDE_SIZE = SECTOR_SIZE * d; 116 | max_qd = atoi(argv[6]); 117 | int FIRST = atoi(argv[7]); 118 | //max_qd = 16; 119 | pattern = atoi(argv[6])/16; 120 | job_queue.set_max(max_qd); 121 | 122 | printf("To run with:\n jump: %d, num_jobs: %d, stride: %d, device_name: %s, write_order: %s, max_qd: %d\n", D, j, d, argv[1], argv[5], max_qd); 123 | 124 | // open raw block device 125 | fd = open(argv[1], O_RDONLY | O_DIRECT); // O_DIRECT 126 | if (fd < 0) { 127 | printf("Raw Device Open failed\n"); 128 | return 1; 129 | } 130 | // context to do async IO 131 | memset(&ctx_, 0, sizeof(ctx_)); 132 | if (io_setup(MAX_COUNT, &ctx_) != 0) { 133 | std::cout << "io_context_t set failed" << std::endl; 134 | exit(1); 135 | } 136 | // generate read trace 137 | // 138 | generate_read_trace(argv[5][0]); 139 | 140 | //parallel reads 141 | 142 | // start the number of threads 143 | pthread_t * thread_pool = (pthread_t *)malloc(sizeof(pthread_t) * j); 144 | 145 | 146 | for (int i = 0; i < j; i++) { 147 | int *arg = (int *)malloc(sizeof(*arg)); 148 | *arg = i; 149 | pthread_create(&thread_pool[i], NULL, eachThread, (void *)arg); 150 | } 151 | 152 | 153 | read_buf = (char *) malloc(sizeof(char) * (STRIDE_SIZE + SECTOR_SIZE)); 154 | int ret = posix_memalign((void **)&read_buf, SECTOR_SIZE, STRIDE_SIZE + SECTOR_SIZE); 155 | sleep(1); 156 | 157 | //as a producer 158 | struct timeval start, end; 159 | gettimeofday(&start, NULL); 160 | int onflight_io = 0; 161 | 162 | timeout.tv_sec = 0; 163 | timeout.tv_nsec = 100; 164 | for (int i = 0; i < num_ios; i++) { 165 | // for each request 166 | onflight_io = 0; 167 | if (argv[5][0] == 'r') { 168 | // parallel reads 169 | for (int k = 0; k < max_qd; k++) { 170 | job_queue.add(read_order[i] + k*SECTOR_SIZE); 171 | } 172 | onflight_io = max_qd; 173 | } else if (argv[5][0] == 'p') { 174 | job_queue.add(FIRST*STRIDE_SIZE); 175 | job_queue.add((FIRST+1)*STRIDE_SIZE + D*SECTOR_SIZE); 176 | onflight_io = 2; 177 | } 178 | // wait until all finished 179 | while (onflight_io > 0) { 180 | int ret = io_getevents(ctx_, 0, MAX_COUNT, events, &timeout); 181 | if (ret < 0) { 182 | std::cout << "Getevents Error" << std::endl; 183 | exit(1); 184 | } 185 | if (ret > 0) { 186 | onflight_io -= ret; 187 | } 188 | } 189 | } 190 | 191 | //wait until all IO finished 192 | while (onflight_io) { 193 | int ret = io_getevents(ctx_, 0, MAX_COUNT, events, &timeout); 194 | if (ret < 0) { 195 | std::cout << "Getevents Error" << std::endl; 196 | exit(1); 197 | } 198 | if (ret > 0) { 199 | onflight_io -= ret; 200 | //std::cout << "=== Get Events("<< ret << "): " << ret << std::endl; 201 | //for (int j = 0; j < ret; j++) { 202 | // std::cout << "=== Complete " << "::" << events[j].obj->aio_fildes << std::endl; 203 | //} 204 | } 205 | } 206 | 207 | 208 | 209 | gettimeofday(&end, NULL); 210 | long time_us = ((end.tv_sec * 1000000 + end.tv_usec) 211 | - (start.tv_sec * 1000000 + start.tv_usec)); 212 | 213 | overall_us += time_us; 214 | // Signal to exit 215 | for (int i = 0; i < j; i++) { 216 | job_queue.add(-1); 217 | } 218 | for (int i = 0; i < j; i++) { 219 | pthread_join(thread_pool[i], NULL); 220 | } 221 | 222 | std::cout << "All IO finished" << std::endl; 223 | printf("Time taken: %ld us, Bandwidth: %f MB/s \n", time_us, (float)((long)STRIDE_SIZE*num_ios)/1024/1024*1000000/time_us); 224 | //printf("Time taken: %ld us, Bandwidth: %f MB/s \n", overall_us/TRIES, (float)((long)STRIDE_SIZE*num_ios)/1024/1024*1000000/(overall_us/TRIES)); 225 | 226 | free(read_buf); 227 | close(fd); 228 | return 0; 229 | } 230 | -------------------------------------------------------------------------------- /random_overwrite: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/sherlockwu/OptaneBench/90ef168cbd79cbfbc57122069b7cd79474bd325c/random_overwrite -------------------------------------------------------------------------------- /random_overwrite.cpp: -------------------------------------------------------------------------------- 1 | #define _GNU_SOURCE 2 | 3 | #include 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | #include 12 | #include 13 | #include 14 | // Note: all pos in sector unless before passed in read/write 15 | int SECTOR_SIZE = 512; 16 | int S = 0; 17 | int n = 0; 18 | 19 | 20 | int main(int argc, char* * argv) { 21 | // identify chunk size 22 | if (argc < 4) { 23 | printf("Wrong parameters: random_overwrite dev_name S(overwritten zone in GB, start from 0) n(req_size in sector) \n"); 24 | return 1; 25 | } 26 | S = atoi(argv[2]); 27 | n = atoi(argv[3]); 28 | 29 | char * write_buf = (char *) malloc(sizeof(char) * (n * SECTOR_SIZE)); 30 | int ret = posix_memalign((void **)&write_buf, SECTOR_SIZE, n*SECTOR_SIZE); 31 | 32 | //printf("To run with:\n zone_size: %d GB, request_size: %d, device_name: %s\n", S, n, argv[1]); 33 | 34 | // open raw block device 35 | int fd = open(argv[1], O_WRONLY | O_DIRECT); 36 | if (fd < 0) { 37 | printf("Raw Device Open failed\n"); 38 | return 1; 39 | } 40 | 41 | //generate_random_overwrite_order(); 42 | std::vector pos_vector; 43 | for (long i = 0; i < (long)(S)*1024*1024*1024; i+=n*SECTOR_SIZE) pos_vector.push_back(i); 44 | std::random_shuffle (pos_vector.begin(), pos_vector.end() ); 45 | std::cout << pos_vector.size() << std::endl; 46 | for (auto it = pos_vector.begin(); it != pos_vector.end(); it++) { 47 | std::cout << *it << std::endl; 48 | int sz = pwrite(fd, write_buf, n * SECTOR_SIZE, *it); 49 | assert(sz == n*SECTOR_SIZE); 50 | } 51 | 52 | 53 | //for (int i = 0; i < S; i+=n) { 54 | // long pos = rand_pos(M); 55 | // int sz = pwrite(fd, write_buf, n * SECTOR_SIZE); 56 | // assert(sz == n*SECTOR_SIZE); 57 | //} 58 | 59 | free(write_buf); 60 | close(fd); 61 | return 0; 62 | } 63 | -------------------------------------------------------------------------------- /setup_cpu.sh: -------------------------------------------------------------------------------- 1 | for CPUFREQ in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; 2 | do [ -f $CPUFREQ ] || continue; echo -n performance > $CPUFREQ; done 3 | 4 | # cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor 5 | 6 | -------------------------------------------------------------------------------- /trim_all.fio: -------------------------------------------------------------------------------- 1 | ; trim all data 2 | 3 | [global] 4 | name=fio-trim 5 | ioengine=libaio 6 | direct=1 7 | thread=1 8 | buffered=0 9 | randrepeat=0 10 | ramp_time=0 11 | norandommap 12 | refill_buffers 13 | log_max_value=1 14 | group_reporting 15 | percentile_list=50.0:99.0 16 | filename=/dev/nvme1n1 17 | 18 | [preconditioning] 19 | stonewall 20 | size=100% 21 | bs=4m 22 | iodepth=16 23 | numjobs=1 24 | rw=trim 25 | 26 | --------------------------------------------------------------------------------