1 /*! \page faq Frequently Asked Questions
3 \htmlinclude .FAQ.doc.toc
5 \section faq_installation Installing the SimGrid library
7 Many people have been asking me questions on how to use SimGrid. Quite
8 often, the questions were not really about SimGrid but on the
9 installation process. This section is intended to help people that are
10 not familiar with compiling C files under UNIX. If you follow these
11 instructions and still have some troubles, drop an e-mail to
12 <simgrid-user@lists.gforge.inria.fr>.
14 \subsection faq_compiling Compiling SimGrid from an archive
16 First of all, you need to download the latest version of SimGrid from
17 <a href="http://gforge.inria.fr/frs/?group_id=12">here</a>.
18 Suppose you have uncompressed SimGrid in some temporary location of
19 your home directory (say <tt>/home/joe/tmp/simgrid-3.0.1 </tt>). The
20 simplest way to use SimGrid is to install it in your home
21 directory. Change your directory to
22 <tt>/home/joe/tmp/simgrid-3.0.1</tt> and type
24 \verbatim./configure --prefix=$HOME
29 If at some point, something fails, check the section "\ref
30 faq_compil_trouble". If it does not help, you can report this problem to the
31 list but, please, avoid sending a laconic mail like "There is a problem. Is it
32 okay?". Send the config.log file which is automatically generated by
33 configure. Try to capture both the standard output and the error output of the
34 <tt>make</tt> command with <tt>script</tt>. There is no way for us to help you
35 without the relevant bits of information.
37 Now, the following directory should have been created :
39 \li <tt>/home/joe/doc/simgrid/html/</tt>
40 \li <tt>/home/joe/lib/</tt>
41 \li <tt>/home/joe/include/</tt>
43 SimGrid is not a binary, it is a library. Both a static and a dynamic
44 version are available. Here is what you can find if you try a <tt>ls
47 \verbatim libsimgrid.a libsimgrid.la libsimgrid.so libsimgrid.so.0 libsimgrid.so.0.0.1
50 Thus, there is two ways to link your program with SimGrid:
51 \li Either you use the static version, e.g
52 \verbatim gcc libsimgrid.a -o MainProgram MainProgram.c
54 In this case, all the SimGrid functions are directly
55 included in <tt>MainProgram</tt> (hence a bigger binary).
56 \li Either you use the dynamic version (the preferred method)
57 \verbatim gcc -lsimgrid -o MainProgram MainProgram.c
59 In this case, the SimGrid functions are not included in
60 <tt>MainProgram</tt> and you need to set your environment
61 variable in such a way that <tt>libsimgrid.so</tt> will be
62 found at runtime. This can be done by adding the following
63 line in your .bashrc (if you use bash and if you have
64 installed the SimGrid libraries in your home directory):
65 \verbatim export LD_LIBRARY_PATH=$HOME/lib/:$LD_LIBRARY_PATH
69 \subsection faq_compiling_cvs Compiling SimGrid from the CVS
71 The project development takes place in the cvs, where all changes are
72 commited when they happen. Then every once in a while, we make sure that the
73 code quality meets our standard and release an archive from the code in the
74 CVS. We afterward go back to the development in the CVS. So, if you need a
75 recently added feature and can afford some little problem with the stability
76 of the lastest features, you may want to use the CVS version instead of a
79 For that, you first need to get the "simgrid" module from
80 <a href="http://gforge.inria.fr/scm/?group_id=12">here</a>.
82 You won't find any <tt>configure</tt> and a few other things
83 (<tt>Makefile.in</tt>'s, documentation, ...) will be missing as well. The
84 reason for that is that all these files have to be regenerated using the
85 latest versions of <tt>autoconf</tt>, <tt>libtool</tt>, <tt>automake</tt>
86 (>1.9) and <tt>doxygen</tt> (>1.4). To generate the <tt>configure</tt> and
87 the <tt>Makefile.in</tt>'s, you just have to launch the <tt>bootstrap</tt>
88 command that resides in the top of the source tree. Then just follow the
89 instructions of Section \ref faq_compiling.
91 We insist on the fact that you really need the latest versions of
92 autoconf, automake and libtool. Doing this step on exotic architectures/systems
93 (i.e. anything different from a recent linux distribution) may be
94 ... uncertain. If you need to compile the CVS version on a machine where all these
95 dependencies are not met, the easiest is to do <tt>make dist</tt> in the CVS
96 dir of another machine where all dependencies are met. It will create an
97 archive you may deploy on other sites just as a regular stable release.
99 In summary, the following commands will checkout the CVS, regenerate the
100 configure script and friends, configure SimGrid and build it.
102 \verbatim cvs -d :pserver:anonymous@scm.gforge.inria.fr:/cvsroot/simgrid login
103 cvs -d :pserver:anonymous@scm.gforge.inria.fr:/cvsroot/simgrid checkout simgrid
106 ./configure --enable-maintainer-mode --prefix=<where to install SimGrid>
109 Then, if you want to install SimGrid on the current box, just do:
110 \verbatim make install \endverbatim
112 If you want to build an snapshot of the CVS to deploy it on another box (for
113 example because the other machine don't have the autotools), do:
114 \verbatim make dist \endverbatim
116 Moreover, you should never call the autotools manually since you must run
117 them in a specific order with specific arguments. Most of the times, the
118 makefiles will automatically call the tools for you. When it's not possible
119 (such as the first time you checkout the CVS), use the ./bootstrap command
120 to call them explicitely.
123 \subsection faq_setting_MSG Setting up your own MSG code
125 Do not build your simulator by modifying the SimGrid examples. Go
126 outside the SimGrid source tree and create your own working directory
127 (say <tt>/home/joe/SimGrid/MyFirstScheduler/</tt>).
129 Suppose your simulation has the following structure (remember it is
130 just an example to illustrate a possible way to compile everything;
131 feel free to organize it as you want).
133 \li <tt>sched.h</tt>: a description of the core of the
134 scheduler (i.e. which functions are can be used by the
135 agents). For example we could find the following functions
136 (master, forwarder, slave).
138 \li <tt>sched.c</tt>: a C file including <tt>sched.h</tt> and
139 implementing the core of the scheduler. Most of these
140 functions use the MSG functions defined in section \ref
143 \li <tt>masterslave.c</tt>: a C file with the main function, i.e.
144 the MSG initialization (MSG_global_init()), the platform
145 creation (e.g. with MSG_create_environment()), the
146 deployment phase (e.g. with MSG_function_register() and
147 MSG_launch_application()) and the call to
150 To compile such a program, we suggest to use the following
151 Makefile. It is a generic Makefile that we have used many times with
152 our students when we teach the C language.
156 masterslave: masterslave.o sched.o
158 INSTALL_PATH = $$HOME
160 PEDANTIC_PARANOID_FREAK = -O0 -Wshadow -Wcast-align \
161 -Waggregate-return -Wmissing-prototypes -Wmissing-declarations \
162 -Wstrict-prototypes -Wmissing-prototypes -Wmissing-declarations \
163 -Wmissing-noreturn -Wredundant-decls -Wnested-externs \
164 -Wpointer-arith -Wwrite-strings -finline-functions
165 REASONABLY_CAREFUL_DUDE = -Wall
166 NO_PRAYER_FOR_THE_WICKED = -w -O2
167 WARNINGS = $(REASONABLY_CAREFUL_DUDE)
168 CFLAGS = -g $(WARNINGS)
170 INCLUDES = -I$(INSTALL_PATH)/include
171 DEFS = -L$(INSTALL_PATH)/lib/
172 LDADD = -lm -lsimgrid
176 $(CC) $(INCLUDES) $(DEFS) $(CFLAGS) $^ $(LIBS) $(LDADD) -o $@
179 $(CC) $(INCLUDES) $(DEFS) $(CFLAGS) -c -o $@ $<
182 rm -f $(BIN_FILES) *.o *~
188 The first two lines indicates what should be build when typing make
189 (<tt>masterslave</tt>) and of which files it is to be made of
190 (<tt>masterslave.o</tt> and <tt>sched.o</tt>). This makefile assumes
191 that you have set up correctly your <tt>LD_LIBRARY_PATH</tt> variable
192 (look, there is a <tt>LDADD = -lm -lsimgrid</tt>). If you prefer using
193 the static version, remove the <tt>-lsimgrid</tt> and add a
194 <tt>$(INSTALL_PATH)/lib/libsimgrid.a</tt> on the next line, right
195 after the <tt>LIBS = </tt>.
197 More generally, if you have never written a Makefile by yourself, type
198 in a terminal : <tt>info make</tt> and read the introduction. The
199 previous example should be enough for a first try but you may want to
200 perform some more complex compilations...
202 \subsection faq_setting_GRAS Setting up your own GRAS code
204 If you use the GRAS interface instead of the MSG one, then previous section
205 is not the better source of information. Instead, you should check the GRAS
206 tutorial in general, and the \ref GRAS_tut_tour_setup in particular.
208 \subsection faq_crosscompile Cross-compiling a Windows DLL of SimGrid from linux
210 At the moment, we do not distribute Windows pre-compiled version of SimGrid
211 because the support for this platform is still experimental. We know that
212 some parts of the GRAS environment do not work, and we think that the others
213 environments (MSG and SD) have good chances to work, but we didn't test
214 ourselves. This section explains how we generate the SimGrid DLL so that you
215 can build it for yourself. First of all, you need to have a version more
216 recent than 3.1 (ie, a CVS version as time of writting).
218 In order to cross-compile the package to windows from linux, you need to
219 install mingw32 (minimalist gnu win32). On Debian, you can do so by
220 installing the packages mingw32 (compiler), mingw32-binutils (linker and
221 so), mingw32-runtime.
223 You can use the VPATH support of configure to compile at the same time for
224 linux and windows without dupplicating the source nor cleaning the tree
225 between each. Just run bootstrap (if you use the CVS) to run the autotools.
226 Then, create a linux and a win directories. Then, type:
227 \verbatim cd linux; ../configure --srcdir=.. <usual configure flags>; make; cd ..
228 cd win; ../configure --srcdir=.. --host=i586-mingw32msvc <flags>; make; cd ..
230 The trick to VPATH builds is to call configure from another directory,
231 passing it an extra --srcdir argument to tell it where all the sources are.
232 It will understand you want to use VPATH. Then, the trick to cross-compile
233 is simply to add a --host argument specifying the target you want to build
234 for. The i586-mingw32msvc string is what you have to pass to use the mingw32
235 environment as distributed in Debian.
237 After that, you can run all make targets from both directories, and test
238 easily that what you change for one arch does not break the other one.
240 It is possible that this VPATH build thing breaks from time to time in the
241 CVS since it's quite fragile, but it's granted to work in any released
242 version. If you experience problems, drop us a mail.
244 Another possible source of issue is that at the moment, building the
245 examples request to use the gras_stub_generator tool, which is a compiled
246 program, not a script. In cross-compilation, you need to cross-execute with
247 wine for example, which is not really pleasant. We are working on this, but
248 in the meanwhile, simply don't build the examples in cross-compilation
249 (<tt>cd src</tt> before running make).
251 Program (cross-)compiled with mingw32 do request an extra DLL at run-time to be
252 usable. For example, if you want to test your build with wine, you should do
253 the following to put this library where wine looks for DLLs.
255 cp /usr/share/doc/mingw32-runtime/mingwm10.dll.gz ~/.wine/c/windows/system/
256 gunzip ~/.wine/c/windows/system/mingwm10.dll.gz
259 The DLL is builded in src/.libs, and installed in the <i>prefix</i>/bin directory
260 when you run make install.
262 If you want to use it in a native project on windows, you need to use
263 simgrid.dll and mingwm10.dll. For each DLL, you need to build .def file
264 under linux (listing the defined symbols), and convert it into a .lib file
265 under windows (specifying this in a way that windows compilers like). To
266 generate the def files, run (under linux):
267 \verbatim echo "LIBRARY libsimgrid-0.dll" > simgrid.def
268 echo EXPORTS >> simgrid.def
269 nm libsimgrid-0.dll | grep ' T _' | sed 's/.* T _//' >> simgrid.def
270 nm libsimgrid-0.dll | grep ' D _' | sed 's/.* D _//' | sed 's/$/ DATA/' >> simgrid.def
272 echo "LIBRARY mingwm10.dll" > mingwm10.def
273 echo EXPORTS >> mingwm10.def
274 nm mingwm10.dll | grep ' T _' | sed 's/.* T _//' >> mingwm10.def
275 nm mingwm10.dll | grep ' D _' | sed 's/.* D _//' | sed 's/$/ DATA/' >> mingwm10.def
278 To create the import .lib files, use the <tt>lib</tt> windows tool (from
279 MSVC) the following way to produce simgrid.lib and mingwm10.lib
280 \verbatim lib /def:simgrid.def
281 lib /def:mingwm10.def
284 If you happen to use Borland C Builder, the right command line is the
285 following (note that you don't need any file.def to get this working).
286 \verbatim implib simgrid.lib libsimgrid-0.dll
287 implib mingwm10.lib mingwm10.dll
290 Then, set the following parameters in Visual C++ 2005:
291 Linker -> Input -> Additional dependencies = simgrid.lib mingwm10.lib
293 Just in case you wonder how to generate a DLL from libtool in another
294 project, we added -no-undefined to any lib*_la_LDFLAGS variables so that
295 libtool accepts to generate a dynamic library under windows. Then, to make
296 it true, we pass any dependencies (such as -lws2 under windows or -lpthread
297 on need) on the linking line. Passing such deps is a good idea anyway so
298 that they get noted in the library itself, avoiding the users to know about
299 our dependencies and put them manually on their compilation line. Then we
300 added the AC_LIBTOOL_WIN32_DLL macro just before AC_PROG_LIBTOOL in the
301 configure.ac. It means that we exported any symbols which need to be.
302 Nowadays, functions get automatically exported, so we don't need to load our
303 header files with tons of __declspec(dllexport) cruft. We only need to do so
304 for data, but there is no public data in SimGrid so we are good.
306 \section faq_simgrid I'm new to SimGrid. I have some questions. Where should I start?
308 You are at the right place... Having a look to these
309 <a href="http://graal.ens-lyon.fr/~alegrand/articles/slides_g5k_simul.pdf">slides</a>
311 <a href="http://graal.ens-lyon.fr/~alegrand/articles/Simgrid-Introduction.pdf">"obsolete" slides</a>)
312 may give you some insights on what SimGrid can help you to do and what
313 are its limitations. Then you definitely should read the \ref
314 MSG_examples. There is also a mailing list: <simgrid-user@lists.gforge.inria.fr>.
316 \subsection faq_interfaces What is the difference between MSG, SimDag, and GRAS? Do they serve the same purpose?
318 It depend on how you define "purpose", I guess ;)
320 They all allow you to build a prototype of application which you can run
321 within the simulator afterward. They all share the same simulation kernel,
322 which is the core of the SimGrid project. They differ by the way you express
325 With SimDag, you express your code as a collection of interdependent
326 parallel tasks. So, in this model, applications can be seen as a DAG of
329 With both GRAS and MSG, your application is seen as a set of communicating
330 processes, exchanging data by the way of messages and performing computation
333 The difference between both is that MSG is somehow easier to use, but GRAS
334 is not limitated to the simulator. Once you're done writing your GRAS code,
335 you can run your code both in the simulator or on a real platform. For this,
336 there is two implementations of the GRAS interface, one for simulation, one
337 for real execution. So, you just have to relink your code to chose one of
340 \subsection faq_generic Building a generic simulator
342 Please read carefully the \ref MSG_examples. You'll find in \ref
343 MSG_ex_master_slave a very simple consisting of a master (that owns a bunch of
344 tasks and distributes them) , some slaves (that process tasks whenever
345 they receive one) and some forwarder agents (that simply pass the
346 tasks they receive to some slaves).
348 \subsection faq_visualization Visualizing the schedule
350 It is sometime convenient to "see" how the agents are behaving. If you
351 like colors, you can use <tt>tools/MSG_visualization/colorize.pl </tt>
352 as a filter to your MSG outputs. It works directly with INFO. Beware,
353 INFO() prints on stderr. Do not forget to redirect if you want to
354 filter (e.g. with bash):
356 ./msg_test small_platform.xml small_deployment.xml 2>&1 | ../../tools/MSG_visualization/colorize.pl
359 We also have a more graphical output. Have a look at MSG_paje_output(). It
360 generates an input to <a href="http://www-id.imag.fr/Logiciels/paje/">Paje</a>.
363 <a href="Paje_MSG_screenshot.jpg"><img src="Paje_MSG_screenshot_thn.jpg"></a>
367 \subsection faq_postmortem_analysis Online/postmortem analysis
369 Vizualization with Paje can be seen as a kind of postmortem
370 analysis. However, as soon as you start playing with big simulations,
371 you'll realize that processing such output is kind of tricky. There is
372 so much generic informations that it is hard to find the information
375 As a matter of fact, loging really depends on simulations (e.g. what
376 kind of events is important...). That is why we do not propose a big
377 dump of your whole simulation (it would slow everything down) but give
378 you neat tools to structure you logs. Have a look at \ref XBT_log. In
379 fact, rather than a post-mortem analysis, you may want to do it on the
380 fly. The process you are running can do whatever you want. Have you
381 thought about adding a global structure where you directly compute the
382 informations that are really important rather than writing everything
383 down and then processing huge files ?
385 \subsection faq_C Argh! Do I really have to code in C ?
387 Up until now, there is no binding for other languages. If you use C++,
388 you should be able to use the SimGrid library as a standard C library
389 and everything should work fine (simply <i>link</i> against this
390 library; recompiling SimGrid with a C++ compiler won't work and it
391 wouldn't help if you could).
393 In fact, the bindings needed to allow one to use SimGrid from Perl,
394 Python, Java, etc. are double-layered. The first layer would allow
395 you to call for example the MSG_task_get_name(task) function while
396 what you really want is a proper object wrapping allowing you to call
397 task->name(). That's the purpose of the second layer. The first one
398 is granted with C++ but can be done with tools like
399 <a href="www.swig.org/">swig</a> for other languages like Perl, Ruby,
400 Python, CAML. None of us really need the second one (which is a bit
401 more demanding and cannot be automatically generated) yet and there is
402 no real point in doing the first one without the second. :)
404 As usual, you're welcome to participate.
406 \section faq_MIA How to ....? Is there a function in the API to simply ....?
408 Here is the deal. The whole SimGrid project (MSG, SURF, GRAS, ...) is
409 meant to be kept as simple and generic as possible. We cannot add
410 functions for everybody's need when these functions can easily be
411 built from the ones already in the API. Most of the time, it is
412 possible and when it was not possible we always have upgraded the API
413 accordingly. When somebody asks us a question like "How to do that ?
414 Is there a function in the API to simply do this ?", we're always glad
415 to answer and help. However if we don't need this code for our own
416 need, there is no chance we're going to write it... it's your job! :)
417 The counterpart to our answers is that once you come up with a neat
418 implementation of this feature (task duplication, RPC, thread
419 synchronization, ...), you should send it to us and we will be glad to
420 add it to the distribution. Thus, other people will take advantage of
421 it (and we don't have to answer this question again and again ;).
423 You'll find in this section a few "Missing In Action" features. Many
424 people have asked about it and we have given hints on how to simply do
425 it with MSG. Feel free to contribute...
427 \subsection faq_MIA_examples I want some more complex examples!
429 Many people have come to ask me a more complex example and each time,
430 they have realized afterward that the basics were in the previous three
433 Of course they have often been needing more complex functions like
434 MSG_process_suspend(), MSG_process_resume() and
435 MSG_process_isSuspended() (to perform synchronization), or
436 MSG_task_Iprobe() and MSG_process_sleep() (to avoid blocking
437 receptions), or even MSG_process_create() (to design asynchronous
438 communications or computations). But the examples are sufficient to
441 We know. We should add some more examples, but not really some more
442 complex ones... We should add some examples that illustrate some other
443 functionalities (like how to simply encode asynchronous
444 communications, RPC, process migrations, thread synchronization, ...)
445 and we will do it when we will have a little bit more time. We have
446 tried to document the examples so that they are understandable. Tell
447 us if something is not clear and once again feel free to participate!
450 \subsection faq_MIA_taskdup Missing in action: Task duplication/replication
452 There is no task duplication in MSG. When you create a task, you can
453 process it or send it somewhere else. As soon as a process has sent
454 this task, he doesn't have this task anymore. It's gone. The receiver
455 process has got the task. However, you could decide upon receiving to
456 create a "copy" of a task but you have to handle by yourself the
457 semantic associated to this "duplication".
459 As we already told, we prefer keeping the API as simple as
460 possible. This kind of feature is rather easy to implement by users
461 and the semantic you associate really depends on people. Having a
462 *generic* task duplication mechanism is not that trivial (in
463 particular because of the data field). That is why I would recommand
464 that you write it by yourself even if I can give you advice on how to
467 You have the following functions to get informations about a task:
468 MSG_task_get_name(), MSG_task_get_compute_duration(),
469 MSG_task_get_remaining_computation(), MSG_task_get_data_size(),
470 and MSG_task_get_data().
472 You could use a dictionnary (#xbt_dict_t) of dynars (#xbt_dynar_t). If
473 you still don't see how to do it, please come back to us...
475 \subsection faq_MIA_asynchronous I want to do asynchronous communications in MSG
477 Up until now, there is no asynchronous communications in MSG. However,
478 you can create as many process as you want so you should be able to do
479 whatever you want... I've written a queue module to help implementing
480 some asynchronous communications at low cost (creating thousands of
481 process only to handle communications may be problematic in term of
482 performance at some point). I'll add it in the distribution asap.
484 \subsection faq_MIA_thread_synchronization I need to synchronize my MSG processes
486 You obviously cannot use pthread_mutexes of pthread_conds. The best
487 thing would be to propose similar structures. Unfortunately, we
488 haven't found time to do it yet. However you can try to play with
489 MSG_process_suspend() and MSG_process_resume(). You can even do some
490 synchronization with fake communications (using MSG_task_get(),
491 MSG_task_put() and MSG_task_Iprobe()).
493 \subsection faq_MIA_host_load Where is the get_host_load function hidden in MSG?
495 There is no such thing because its semantic wouldn't be really
496 clear. Of course, it is something about the amount of host throughput,
497 but there is as many definition of "host load" as people asking for
498 this function. First, you have to remember that resource availability
499 may vary over time, which make any load notion harder to define.
501 It may be instantaneous value or an average one. Moreover it may be only the
502 power of the computer, or may take the background load into account, or may
503 even take the currently running tasks into account. In some SURF models,
504 communications have an influence on computational power. Should it be taken
507 First of all, it's near to impossible to predict the load beforehands in the
508 simulator since it depends on too much parameters (background load
509 variation, bandwidth sharing algorithmic complexity) some of them even being
510 not known beforehands (other task starting at the same time). So, getting
511 this information is really hard (just like in real life). It's not just that
512 we want MSG to be as painful as real life. But as it is in some way
513 realistic, we face some of the same problems as we would face in real life.
515 How would you do it for real? The most common option is to use something
516 like NWS that performs active probes. The best solution is probably to do
517 the same within MSG, as in next code snippet. It is very close from what you
518 would have to do out of the simulator, and thus gives you information that
519 you could also get in real settings to not hinder the realism of your
523 double get_host_load() {
524 m_task_t task = MSG_task_create("test", 0.001, 0, NULL);
525 double date = MSG_get_clock();
527 MSG_task_execute(task);
528 date = MSG_get_clock() - date;
529 MSG_task_destroy(task);
534 Of course, it may not match your personal definition of "host load". In this
535 case, please detail what you mean on the mailing list, and we will extend
536 this FAQ section to fit your taste if possible.
538 \subsection faq_MIA_communication_time How can I get the *real* communication time ?
540 Communications are synchronous and thus if you simply get the time
541 before and after a communication, you'll only get the transmission
542 time and the time spent to really communicate (it will also take into
543 account the time spent waiting for the other party to be
544 ready). However, getting the *real* communication time is not really
545 hard either. The following solution is a good starting point.
550 m_task_t task = MSG_task_create("Task", task_comp_size, task_comm_size,
551 calloc(1,sizeof(double)));
552 *((double*) task->data) = MSG_get_clock();
553 MSG_task_put(task, slaves[i % slaves_count], PORT_22);
554 INFO0("Send completed");
559 m_task_t task = NULL;
562 time1 = MSG_get_clock();
563 a = MSG_task_get(&(task), PORT_22);
564 time2 = MSG_get_clock();
565 if(time1<*((double *)task->data))
566 time1 = *((double *) task->data);
567 INFO1("Communication time : \"%f\" ", time2-time1);
569 MSG_task_destroy(task);
574 \subsection faq_MIA_batch_scheduler Is there a native support for batch schedulers in SimGrid ?
576 No, there is no native support for batch schedulers and none is
577 planned because this is a very specific need (and doing it in a
578 generic way is thus very hard). However some people have implemented
579 their own batch schedulers. Vincent Garonne wrote one during his PhD
580 and put his code in the contrib directory of our CVS so that other can
581 keep working on it. You may find inspinring ideas in it.
583 \subsection faq_MIA_checkpointing I need a checkpointing thing
585 Actually, it depends on whether you want to checkpoint the simulation, or to
586 simulate checkpoints.
588 The first one could help if your simulation is a long standing process you
589 want to keep running even on hardware issues. It could also help to
590 <i>rewind</i> the simulation by jumping sometimes on an old checkpoint to
591 cancel recent calculations.\n
592 Unfortunately, such thing will probably never exist in SG. One would have to
593 duplicate all data structures because doing a rewind at the simulator level
594 is very very hard (not talking about the malloc free operations that might
595 have been done in between). Instead, you may be interested in the Libckpt
596 library (http://www.cs.utk.edu/~plank/plank/www/libckpt.html). This is the
597 checkpointing solution used in the condor project, for example. It makes it
598 easy to create checkpoints (at the OS level, creating something like core
599 files), and rerunning them on need.
601 If you want to simulate checkpoints instead, it means that you want the
602 state of an executing task (in particular, the progress made towards
603 completion) to be saved somewhere. So if a host (and the task executing on
604 it) fails (cf. #MSG_HOST_FAILURE), then the task can be restarted
605 from the last checkpoint.\n
607 Actually, such a thing does not exists in SimGrid either, but it's just
608 because we don't think it is fundamental and it may be done in the user code
609 at relatively low cost. You could for example use a watcher that
610 periodically get the remaining amount of things to do (using
611 MSG_task_get_remaining_computation()), or fragment the task in smaller
614 \section faq_SG Where has SG disappeared?!?
616 OK, it's time to explain what's happening to the SimGrid project. Let's
617 start with a little bit of history.
619 * Historically, SimGrid was a low-level toolkit for scheduling with
620 classical models such as DAGs. That was SimGrid v.1.* aka SG, written
621 by Henri Casanova. I (Arnaud) had been using it in its earliest
622 versions during an internship at UCSD.
624 Then we have realized that encoding distributed algorithm in SG was a
627 * So we have built MSG on top of SG and have released SimGrid v.2.*. MSG
628 offered a very basic API to encode a distributed application easily.
629 However encoding MSG on top of SG was not really convenient and did not
630 use the DAG part since the control of the task synchronization was done
631 on top of MSG and no more in SG. We have been playing a little bit with
632 MSG. We have realized that:
634 \li 1) the platform modeling was quite flexible and could be "almost"
635 automated (e.g. using random generator and post-annotations);
637 \li 2) SG was the bottleneck because of the way we were using
638 it. We needed to simulate concurrent transfers, complex load
639 sharing mechanisms. Many optimizations (e.g. trace integration)
640 were totally inefficient when combined with MSG and made extending SG
641 to implement new sharing policies, parallel tasks models, or failures
642 (many people were asking for these kind of features) a real pain;
644 \li 3) the application modeling was not really easy. Even though the
645 application modeling depends on people's applications, we thought
646 we could improve things here. One of our target here was realistic
647 distributed applications ranging from computer sensor networks like
648 the NWS to peer-to-peer applications;
650 * So we have been planning mainly two things for SimGrid 3:
652 \li 1) I have proposed to get rid of SG and to re-implement a new kernel
653 that would be faster and more flexible. That is what I did in the
654 end of 2004: SURF. SURF is based on a fast max-min linear solver
655 using O(1) data-structures. I have quickly replaced SG by SURF in
656 MSG and the result has been that on the MSG example, the new
657 version was more than 10 times faster while we had gain a lot of
658 flexibility. I think I could still easily make MSG faster but I
659 have to work on MSG now (e.g. using some of the O(1)
660 data-structures I've been using to build SURF) since it has become
661 the bottleneck. Some MSG functions have been removed from the API
662 but they were mainly intended to build the platform by hand (they
663 had appeared in the earliest versions of MSG) and were therefore
664 not useful anymore since we are providing a complete mechanism to
665 automatically build the platform and deploy the agents on it.;
667 \li 2) GRAS is a new project Martin and I have come up with. The idea is
668 to have a programming environment that let you program real
669 distributed applications while letting you the ability to run it in
670 the simulator without having to change the slightest line of your
671 code. From the simulation point of view, GRAS performs the
672 application modeling automatically... Up until now, GRAS works on
673 top MSG for historical reasons but I'm going to make it work
674 directly on top of SURF so that it can use all the flex and the
675 speed provided by SURF.
677 Those two things are working, but we want to make everything as clean as
678 possible before releasing SimGrid v.3.
680 So what about those nice DAGs we used to have in SimGrid v.1.? They're
681 not anymore in SimGrid v.3. At least not in their original form... Let
682 me recall you the way SimGrid 3 is organized:
696 XBT is our tool box and now, you should have an idea of what the other
697 ones are. As you can see, the primitive SG is not here
698 anymore. However we have written a brand new and cleaner API for this
699 purpose: \ref SD_API. It is built directly on top of SURF and provides
700 an API rather close to the old SG:
703 ______________________
705 |____________________|
706 | | MSG | GRAS | SD |
707 | -------------------|
709 | -------------------|
711 ----------------------
714 The nice thing is that, as it is writen on top of SURF, it seamlessly
715 support DAG of parallel tasks as well as complex communications
716 patterns. Some old codes using SG are currently under rewrite using
717 \ref SD_API to check that all needful functions are provided.
719 \subsection faq_SG_DAG How to implement a distributed dynamic scheduler of DAGs.
721 Distributed is somehow "contagious". If you start making distributed
722 decisions, there is no way to handle DAGs directly anymore (unless I
723 am missing something). You have to encode your DAGs in term of
724 communicating process to make the whole scheduling process
725 distributed. Here is an example of how you could do that. Assume T1
726 has to be done before T2.
729 int your_agent(int argc, char *argv[] {
731 T1 = MSG_task_create(...);
732 T2 = MSG_task_create(...);
736 if(cond) MSG_task_execute(T1);
738 if((MSG_task_get_remaining_computation(T1)=0.0) && (you_re_in_a_good_mood))
741 /* do something else */
747 If you decide that the distributed part is not that much important and that
748 DAG is really the level of abstraction you want to work with, then you should
749 give a try to \ref SD_API.
751 \section faq_dynamic Dynamic resources and platform building
753 \subsection faq_platform Building a realistic platform
755 We can speak more than an hour on this subject and we still do not have
756 the right answer, just some ideas. You can read the following
757 <a href="http://graal.ens-lyon.fr/~alegrand/articles/Simgrid-Introduction.pdf">slides</a>.
758 It may give you some hints. You can also have a look at the
759 <tt>tools/platform_generation/</tt> directory. There is a perl-script
760 we use to annotate a Tiers generated platform.
762 \subsection faq_SURF_dynamic How can I have variable resource availability?
764 A nice feature of SimGrid is that it enables you to seamlessly have
765 resources whose availability change over time. When you build a
766 platform, you generally declare CPUs like that:
769 <cpu name="Cpu A" power="100.00"/>
772 If you want the availability of "CPU A" to change over time, the only
773 thing you have to do is change this definition like that:
776 <cpu name="Cpu A" power="100.00" availability_file="trace_A.txt" state_file="trace_A_failure.txt"/>
779 For CPUs, availability files are expressed in fraction of available
780 power. Let's have a look at what "trace_A.txt" may look like:
789 At time 0, our CPU will deliver 100 flop/s. At time 11.0, it will
790 deliver only 50 flop/s until time 20.0 where it will will start
791 delivering 90 flop/s. Last at time 21.0 (20.0 plus the periodicity
792 1.0), we'll be back to the beginning and it will deliver 100 flop/s.
794 Now let's look at the state file:
801 A negative value means "off" while a positive one means "on". At time
802 1.0, the CPU is on. At time 1.0, it is turned off and at time 2.0, it
803 is turned on again until time 12 (2.0 plus the periodicity 10.0). It
804 will be turned on again at time 13.0 until time 23.0, and so on.
806 Now, let's look how the same kind of thing can be done for network
807 links. A usual declaration looks like:
810 <network_link name="LinkA" bandwidth="10.0" latency="0.2"/>
813 You have at your disposal the following options: bandwidth_file,
814 latency_file and state_file. The only difference with CPUs is that
815 bandwidth_file and latency_file do not express fraction of available
816 power but are expressed directly in bytes per seconds and seconds.
818 \subsection faq_flexml_bypassing How can I have some C functions do what the platform file does?
820 So you want to bypass the XML files parser, uh? Maybe doin some parameter
821 sweep experiments on your simulations or so? This is possible, but it's not
822 really easy. Here is how it goes.
824 For this, you have to first remember that the XML parsing in SimGrid is done
825 using a tool called FleXML. Given a DTD, this gives a flex-based parser. If
826 you want to bypass the parser, you need to provide some code mimicking what
827 it does and replacing it in its interactions with the SURF code. So, let's
828 have a look at these interactions.
830 FleXML parser are close to classical SAX parsers. It means that a
831 well-formed SimGrid platform XML file might result in the following
834 - start "platform_description"
835 - start "cpu" with attributes name="host1" power="1.0"
837 - start "cpu" with attributes name="host2" power="2.0"
839 - start "network_link" with ...
841 - start "route" with ...
843 - start "route" with ...
845 - end "platform_description"
847 The communication from the parser to the SURF code uses two means:
848 Attributes get copied into some global variables, and a surf-provided
849 function gets called by the parser for each event. For example, the event
850 - start "cpu" with attributes name="host1" power="1.0"
852 let the parser do the equivalent of:
854 strcpy("host1",A_cpu_name);
859 In SURF, we attach callbacks to the different events by initializing the
860 pointer functions to some the right surf functions. Example in
861 workstation_KCCFLN05.c (surf_parse_open() ends up calling surf_parse()):
863 // Building the routes
864 surf_parse_reset_parser();
865 STag_route_fun=parse_route_set_endpoints;
866 ETag_route_element_fun=parse_route_elem;
867 ETag_route_fun=parse_route_set_route;
868 surf_parse_open(file);
869 xbt_assert1((!surf_parse()),"Parse error in %s",file);
873 So, to bypass the FleXML parser, you need to write your own version of the
874 surf_parse function, which should do the following:
875 - Call the corresponding STag_<tag>_fun function to simulate tag start
876 - Fill the A_<tag>_<attribute> variables with the wanted values
877 - Call the corresponding ETag_<tag>_fun function to simulate tag end
878 - (do the same for the next set of values, and loop)
880 Then, tell SimGrid that you want to use your own "parser" instead of the stock one:
882 surf_parse = surf_parse_bypass;
883 MSG_create_environment(NULL);
886 An example of this trick is distributed in the file examples/msg/msg_test_surfxml_bypassed.c
888 \section faq_limits Pushing the limits
890 \subsection faq_context_1000 I want thousands of simulated processes
892 SimGrid can use either pthreads library or the UNIX98 contextes. On most
893 systems, the number of pthreads is limited and then your simulation may be
894 limited for a stupid reason. This is especially true with the current linux
895 pthreads, and I cannot get more than 2000 simulated processes with pthreads
896 on my box. The UNIX98 contexts allow me to raise the limit to 25,000
897 simulated processes on my laptop.
899 The <tt>--with-context</tt> option of the <tt>./configure</tt> script allows
900 you to choose between UNIX98 contextes (<tt>--with-context=ucontext</tt>)
901 and the pthread version ( (<tt>--with-context=pthread</tt>). The default
902 value is ucontext when the script detect a working UNIX98 context
903 implementation. On Windows boxes, the provided value is discarded and an
904 adapted version is picked up.
906 We experienced some issues with contextes on some rare systems (solaris 8
907 and lower or old alpha linuxes comes to mind). The main problem is that the
908 configure script detect the contextes as being functional when it's not
909 true. If you happen to use such a system, switch manually to the pthread
910 version, and provide us with a good patch for the configure script so that
911 it is done automatically ;)
913 \subsection faq_context_10000 I want hundred thousands of simulated processes
915 As explained above, SimGrid can use UNIX98 contextes to represent and handle
916 the simulated processes. Thanks to this, the main limitation to the number
917 of simulated processes becomes the available memory.
919 Here are some tricks I had to use in order to run a token ring between
920 25,000 processes on my laptop (1Gb memory, 1.5Gb swap).
922 - First of all, make sure your code runs for a few hundreds processes
923 before trying to push the limit. Make sure it's valgrind-clean, ie that
924 valgrind does not report neither memory error nor memory leaks. Indeed,
925 numerous simulated processes result in *fat* simulation hindering debugging.
927 - It was really boring to write 25,000 entries in the deployment file, so I wrote
928 a little script <tt>examples/gras/tokenS/make_deployment.pl</tt>, which you may
929 want to adapt to your case. You could also think about hijacking
930 the SURFXML parser (have look at \ref faq_flexml_bypassing).
932 - The deployment file became quite big, so I had to do what is in the FAQ
933 entry \ref faq_flexml_limit
935 - Each UNIX98 context has its own stack entry. As debugging this is quite
936 hairly, the default value is a bit overestimated so that user don't get
937 into trouble about this. You want to tune this size to increse the number
938 of processes. This is the <tt>STACK_SIZE</tt> define in
939 <tt>src/xbt/context_private.h</tt>, which is 128kb by default.
940 Reduce this as much as you can, but be warned that if this value is too
941 low, you'll get a segfault. The token ring example, which is quite simple,
942 runs with 40kb stacks.
944 \section faq_troubleshooting Troubleshooting
946 \subsection faq_compil_trouble ./configure fails!
948 We now only one reason for the configure to fail:
950 - <b>You are using a borken build environment</b>\n
951 If symptom is that configure complains about gcc not being able to build
952 executables, you are probably missing the libc6-dev package. Damn Ubuntu.
954 If you experience other kind of issue, please get in touch with us. We are
955 always interested in improving our portability to new systems.
957 \subsection faq_distcheck_fails Dude! "make check" fails on my machine!
959 Don't assume we never run this target, because we do. Really. Promise!
961 There is several reasons which may cause the make check to fail on your
964 - <b>You are using a borken libc (probably concerning the contextes)</b>.\n
965 The symptom is that the "make check" fails within the examples/msg directory.\n
966 By default, SimGrid uses something called ucontexts. This is part of the
967 libc, but it's quite undertested. For example, some (old) versions of the
968 glibc on alpha do not implement these functions, but provide the stubs
969 (which return ENOSYS: not implemented). It fools our detection mecanism
970 and leads to segfaults.\n
971 On some x86_64, the pointer to function is stored into a integer, but int
972 are 32bits only on this arch while pointers are 64bits. Our detection
973 mecanism also fails to detect the problem, which leads to segfaults.\n
974 In both cases, there is not much we can do to fix the bug. We are working
975 on a workaround for x86_64 machines, but in the meanwhile, you can
976 compile with --with-context=pthread to avoid ucontext completely. You'll
977 be a bit more limitated in the number of simulated processes you can start
978 concurently, but 5000 processes is still enough for most purposes, isn't
980 This limitation is the reason why we insist on using this piece of ...
981 software even if it's so troublesome.\n
982 <b>=> use --with-pthread on AMD64 architecture that do not have an
983 ultra-recent libc.</b>
985 - <b>There is a bug in SimGrid we aren't aware of</b>.\n
986 If none of the above apply, please drop us a mail on the mailing list so
987 that we can check it out.
989 \subsection faq_longjmp longjmp madness in valgrind
991 This is when valgrind starts complaining about longjmp things, just like:
993 \verbatim ==21434== Conditional jump or move depends on uninitialised value(s)
994 ==21434== at 0x420DBE5: longjmp (longjmp.c:33)
996 ==21434== Use of uninitialised value of size 4
997 ==21434== at 0x420DC3A: __longjmp (__longjmp.S:48)
1000 or even when it reports scary things like:
1002 \verbatim ==24023== Warning: client switching stacks? SP change: 0xBE3FF618 --> 0xBE7FF710
1003 x86->IR: unhandled instruction bytes: 0xF4 0xC7 0x83 0xD0
1004 ==24023== to suppress, use: --max-stackframe=4194552 or greater
1005 ==24023== Your program just tried to execute an instruction that Valgrind
1006 ==24023== did not recognise. There are two possible reasons for this.
1007 ==24023== 1. Your program has a bug and erroneously jumped to a non-code
1008 ==24023== location. If you are running Memcheck and you just saw a
1009 ==24023== warning about a bad jump, it's probably your program's fault.
1010 ==24023== 2. The instruction is legitimate but Valgrind doesn't handle it,
1011 ==24023== i.e. it's Valgrind's fault. If you think this is the case or
1012 ==24023== you are not sure, please let us know.
1013 ==24023== Either way, Valgrind will now raise a SIGILL signal which will
1014 ==24023== probably kill your program.
1016 ==24023== Process terminating with default action of signal 4 (SIGILL)
1017 ==24023== Illegal opcode at address 0x420D234
1018 ==24023== at 0x420D234: abort (abort.c:124)
1021 This is the sign that you didn't used the exception mecanism well. Most
1022 probably, you have a <tt>return;</tt> somewhere within a <tt>TRY{}</tt>
1023 block. This is <b>evil</b>, and you must not do this. Did you read the section
1026 \subsection faq_valgrind Valgrind spits tons of errors!
1028 It may happen that valgrind, the memory debugger beloved by any decent C
1029 programmer, spits tons of warnings like the following :
1030 \verbatim ==8414== Conditional jump or move depends on uninitialised value(s)
1031 ==8414== at 0x400882D: (within /lib/ld-2.3.6.so)
1032 ==8414== by 0x414EDE9: (within /lib/tls/i686/cmov/libc-2.3.6.so)
1033 ==8414== by 0x400B105: (within /lib/ld-2.3.6.so)
1034 ==8414== by 0x414F937: _dl_open (in /lib/tls/i686/cmov/libc-2.3.6.so)
1035 ==8414== by 0x4150F4C: (within /lib/tls/i686/cmov/libc-2.3.6.so)
1036 ==8414== by 0x400B105: (within /lib/ld-2.3.6.so)
1037 ==8414== by 0x415102D: __libc_dlopen_mode (in /lib/tls/i686/cmov/libc-2.3.6.so)
1038 ==8414== by 0x412D6B9: backtrace (in /lib/tls/i686/cmov/libc-2.3.6.so)
1039 ==8414== by 0x8076446: xbt_dictelm_get_ext (dict_elm.c:714)
1040 ==8414== by 0x80764C1: xbt_dictelm_get (dict_elm.c:732)
1041 ==8414== by 0x8079010: xbt_cfg_register (config.c:208)
1042 ==8414== by 0x806821B: MSG_config (msg_config.c:42)
1045 This problem is somewhere in the libc when using the backtraces and there is
1046 very few things we can do ourselves to fix it. Instead, here is how to tell
1047 valgrind to ignore the error. Add the following to your ~/.valgrind.supp (or
1048 create this file on need). Make sure to change the obj line according to
1049 your personnal mileage (change 2.3.6 to the actual version you are using,
1050 which you can retrieve with a simple "ls /lib/ld*.so").
1053 name: Backtrace madness
1055 obj:/lib/ld-2.3.6.so
1060 fun:__libc_dlopen_mode
1063 Then, you have to specify valgrind to use this suppression file by passing
1064 the <tt>--suppressions=$HOME/.valgrind.supp</tt> option on the command line.
1065 You can also add the following to your ~/.bashrc so that it gets passed
1066 automatically. Actually, it passes a bit more options to valgrind, and this
1067 happen to be my personnal settings. Check the valgrind documentation for
1070 \verbatim export VALGRIND_OPTS="--leak-check=yes --leak-resolution=high --num-callers=40 --tool=memcheck --suppressions=$HOME/.valgrind.supp" \endverbatim
1072 \subsection faq_flexml_limit I get the message "surf_parse_lex: Assertion `next<limit' failed."
1074 This is because your platform file is too big for the parser.
1076 Actually, the message comes directly from FleXML, the technology on top of
1077 which the parser is built. FleXML has the bad idea of fetching the whole
1078 document in memory before parsing it. And moreover, the memory buffer size
1079 must be determinded at compilation time.
1081 We use a value which seems big enough for our need without bloating the
1082 simulators footprints. But of course your mileage may vary. In this case,
1083 just edit src/surf/surfxml.l modify the definition of
1084 FLEXML_BUFFERSTACKSIZE. E.g.
1087 #define FLEXML_BUFFERSTACKSIZE 1000000000
1090 Then recompile and everything should be fine, provided that your version of
1091 Flex is recent enough (>= 2.5.31). If not the compilation process should
1094 A while ago, we worked on FleXML to reduce a bit its memory consumtion, but
1095 these issues remain. There is two things we should do:
1097 - use a dynamic buffer instead of a static one so that the only limit
1098 becomes your memory, not a stupid constant fixed at compilation time
1099 (maybe not so difficult).
1100 - change the parser so that it does not need to get the whole file in
1101 memory before parsing
1102 (seems quite difficult, but I'm a complete newbe wrt flex stuff).
1104 These are changes to FleXML itself, not SimGrid. But since we kinda hijacked
1105 the development of FleXML, I can grant you that any patches would be really
1106 welcome and quickly integrated.
1108 <b>Update:</b> A new version of FleXML (1.7) was released. Most of the work
1109 was done by William Dowling, who use it in his own work. The good point is
1110 that it now use a dynamic buffer, and that the memory usage was greatly
1111 improved. The downside is that William also changed some things internally,
1112 and it breaks the hack we devised to bypass the parser, as explained in
1113 \ref faq_flexml_bypassing. Indeed, this is not a classical usage of the
1114 parser, and Will didn't imagine that we may have used (and even documented)
1115 such a crude usage of FleXML. So, we now have to repare the bypassing
1116 functionnality to use the lastest FleXML version and fix the memory usage in
1119 \subsection faq_gras_transport GRAS spits networking error messages
1121 Gras, on real platforms, naturally use regular sockets to communicate. They
1122 are deeply hiden in the gras abstraction, but when things go wrong, you may
1123 get some weird error messages. Here are some example, with the probable
1126 - <b>Transport endpoint is not connected</b>: several processes try to open
1127 a server socket on the same port number of the same machine. This is
1128 naturally bad and each process should pick its own port number for this.\n
1129 Maybe, you just have some processes remaining from a previous experiment
1131 Killing them may help, but again if you kill -KILL them, you'll have to
1132 wait for a while: they didn't close there sockets properly and the system
1133 needs a while to notice that this port is free again.
1135 - <b>Socket closed by remote side</b>: if the remote process is not
1136 supposed to close the socket at this point, it may be dead.
1138 - <b>Connection reset by peer</b>: I found this on internet about this
1139 error. I think it's what's happening here, too:\n
1140 <i>This basically means that a network error occurred while the client was
1141 receiving data from the server. But what is really happening is that the
1142 server actually accepts the connection, processes the request, and sends
1143 a reply to the client. However, when the server closes the socket, the
1144 client believes that the connection has been terminated abnormally
1145 because the socket implementation sends a TCP reset segment telling the
1146 client to throw away the data and report an error.\n
1147 Sometimes, this problem is caused by not properly closing the
1148 input/output streams and the socket connection. Make sure you close the
1149 input/output streams and socket connection properly. If everything is
1150 closed properly, however, and the problem persists, you can work around
1151 it by adding a one-second sleep before closing the streams and the
1152 socket. This technique, however, is not reliable and may not work on all
1154 Since GRAS sockets are closed properly (repeat after me: there is no bug
1155 in GRAS), it is either that you are closing your sockets on server side
1156 before the client get a chance to read them (use gras_os_sleep() to delay
1157 the server), or the server died awfully before the client got the data.
1159 \subsection faq_deadlock There is a deadlock !!!
1161 Unfortunately, we cannot debug every code written in SimGrid. We
1162 furthermore believe that the framework provides ways enough
1163 information to debug such informations yourself. If the textual output
1164 is not enough, Make sure to check the \ref faq_visualization FAQ entry to see
1165 how to get a graphical one.
1167 Now, if you come up with a really simple example that deadlocks and
1168 you're absolutely convinced that it should not, you can ask on the
1169 list. Just be aware that you'll be severely punished if the mistake is
1170 on your side... We have plenty of FAQ entries to redact and new
1171 features to implement for the impenitents! ;)
1173 \subsection faq_big_fat_warning A BIG FAT WARNING is reported telling me that my platform and deployment files are too old.
1175 We have decided to change the units in SimGrid. Now we use Bytes, Flops and
1176 seconds instead of MBytes, MFlops and seconds... Units should be updated
1177 accordingly and the version of platform_description should be set to a
1178 valuer greater than 1:
1180 <platform_description version="1">
1182 You should try to use the surfxml_update.pl script that can be found
1183 <a href="http://gforge.inria.fr/plugins/scmcvs/cvsweb.php/contrib/platform_generation/?cvsroot=cvsroot%2Fsimgrid">here</a>.
1185 \subsection faq_surf_network_latency I get weird timings when I play with the latencies.
1187 OK, first of all, remember that units should be Bytes, Flops and
1188 Seconds. If you don't use such units, some SimGrid constants (e.g. the
1189 SG_TCP_CTE_GAMMA constant used in most network models) won't have the
1190 right unit and you'll end up with weird results.
1192 Here is what happens with a single transfer of size L on a link
1193 (bw,lat) when nothing else happens.
1196 0-----lat--------------------------------------------------t
1197 |-----|**** real_bw =min(bw,SG_TCP_CTE_GAMMA/(2*lat)) *****|
1200 In more complex situations, this min is the solution of a complex
1201 max-min linear system. Have a look
1202 <a href="http://lists.gforge.inria.fr/pipermail/simgrid-devel/2006-April/thread.html">here</a>
1203 and read the two threads "Bug in SURF?" and "Surf bug not
1204 fixed?". You'll have a few other examples of such computations. You
1205 can also read "A Network Model for Simulation of Grid Application" by
1206 Henri Casanova and Loris Marchal to have all the details. The fact
1207 that the real_bw is smaller than bw is easy to understand. The fact
1208 that real_bw is smaller than SG_TCP_CTE_GAMMA/(2*lat) is due to the
1209 window-based congestion mechanism of TCP. With TCP, you can't exploit
1210 your huge network capacity if you don't have a good round-trip-time
1211 because of the acks...
1213 Anyway, what you get is t=lat + L/min(bw,SG_TCP_CTE_GAMMA/(2*lat)).
1215 * if I you set (bw,lat)=(100 000 000, 0.00001), you get t = 1.00001 (you fully
1217 * if I you set (bw,lat)=(100 000 000, 0.0001), you get t = 1.0001 (you're on the
1219 * if I you set (bw,lat)=(100 000 000, 0.001), you get t = 10.001 (ouch!)
1221 This bound on the effective bandwidth of a flow is not the only thing
1222 that may make your result be unexpected. For example, two flows
1223 competing on a saturated link receive an amount of bandwidth inversely
1224 proportional to their round trip time.
1226 \subsection faq_bugrepport So I've found a bug in SimGrid. How to report it?
1228 We do our best to make sure to hammer away any bugs of SimGrid, but this is
1229 still an academic project so please be patient if/when you find bugs in it.
1230 If you do, the best solution is to drop an email either on the simgrid-user
1231 or the simgrid-devel mailing list and explain us about the issue. You can
1232 also decide to open a formal bug report using the
1233 <a href="https://gforge.inria.fr/tracker/?atid=165&group_id=12&func=browse">relevant
1234 interface</a>. You need to login on the server to get the ability to submit
1237 We will do our best to solve any problem repported, but you need to help us
1238 finding the issue. Just telling "it segfault" isn't enough. Telling "It
1239 segfaults when running the attached simulator" doesn't really help either.
1240 You may find the following article interesting to see how to repport
1241 informative bug repports:
1242 http://www.chiark.greenend.org.uk/~sgtatham/bugs.html (it is not SimGrid
1243 specific at all, but it's full of good advices).
1245 \author Arnaud Legrand (arnaud.legrand::imag.fr)
1246 \author Martin Quinson (martin.quinson::loria.fr)