5 <img align=center src="simgrid_logo.png" alt="SimGrid"><br>
9 \section quick Quick start
11 SimGrid is a toolkit that provides core functionalities for the simulation
12 of distributed applications in heterogeneous distributed environments.
13 The specific goal of the project is to facilitate research in the area of
14 distributed and parallel application scheduling on distributed computing
15 platforms ranging from simple network of workstations to Computational
18 \subsection quick_dl Getting and installing the software
20 - The official webpage is <a href="http://simgrid.gforge.inria.fr/">simgrid.gforge.inria.fr</a>.
21 - The development webpage is <a href="https://gforge.inria.fr/projects/simgrid//">gforge.inria.fr/projects/simgrid</a>.
22 - The user mailing list is <simgrid-user@lists.gforge.inria.fr>
23 - The SimGrid software package can be downloaded from <a href="https://gforge.inria.fr/frs/?group_id=12">here</a>.
25 To compile and install it, simply type the following. If you are not
26 familiar with compiling C files under UNIX and using libraries, please check
27 the \ref faq. SimGrid also works under Windows, but we do not distribute any
28 pre-compiled binaries [yet].
30 \verbatim $ ./configure
33 # make install\endverbatim
35 \subsection quick_more More information
37 The API is described <a href="API/html/modules.html">here</a> while
38 <a href="examples/html/modules.html">this page</a> presents some example of
41 For more information about the SimGrid toolkit, please simply keep reading
42 this page. It is organized as follow:
44 - \ref overview: Presentation of the toolkit, of each of its components
45 and of their interactions.
46 - \ref people: Who is behind this project.
47 - \ref publications: Some articles providing more details about the
48 SimGrid toolkit or using and validating it.
52 \section overview Overview of the toolkit components
54 As depicted by the following schema, the SimGrid toolkit is basically
61 \htmlinclude simgrid_modules.map
63 <br><b>Relationships between the SimGrid components</b>
67 \subsection overview_fondation Basement layer
69 The basement of the whole toolkit is constituted by the <b>\ref XBT_API
70 (eXtended Bundle of Tools)</b>.
72 It is a portable library providing some grounding features such as \ref
73 XBT_log, \ref XBT_error and \ref XBT_config. XBT also encompass the
74 following convenient datastructures: \ref XBT_dynar, \ref XBT_fifo, \ref
75 XBT_dict, \ref XBT_heap, \ref XBT_set and \ref XBT_swag.
77 See the \ref XBT_API section for more details.
79 \subsection overview_kernel Simulation kernel layer
81 The core functionnalities to simulate a virtual platform are provided by a
82 module called <b>\ref SURF_API</b> ("that's historical, my friend"). It is
83 very low-level and is not intended to be used as such by end-users. Instead,
84 it serve as a basis for the higher level layer.
86 SURF main features are a fast max-min linear solver and the ability to
87 change transparently the model used to describe the platform. This greatly
88 eases the comparison of the several models existing in the litterature.
90 See the \ref SURF_API section for more details.
92 \subsection overview_envs Programmation environments layer
94 This simulation kernel is used to build several programmation environments.
95 Each of them target a specific audiance and constitute a different paradigm.
96 To choose which of them you want to use, you have to think about what you
97 want to do and what would be the result of your work.
99 - If you want to study a theoritical problem and compare several
100 heuristics, you probably want to try <b>\ref MSG_API</b> (yet another
101 historical name). It was designed exactly to that extend and should allow
102 you to build easily rather realistic multi-agents simulation. Yet,
103 realism is not the main goal of this environment and the most annoying
104 technical issues of real platforms are masked here. Check the \ref
105 MSG_API section for more information.
107 - If you want to study the behaviour of a MPI application using emulation
108 technics, you should have a look at the <b>\ref SMPI_API</b> (Simulated
109 MPI) programming environment. Unfortunately, this work is still underway.
110 Check the \ref SMPI_API section for more information.
112 - If you want to develop a real distributed application, then you may find
113 <b>\ref GRAS_API</b> (Grid Reality And Simulation) useful. This is an API
114 for the realization of distributed applications.
116 Moreover, there is two implementations of this API: one on top of the
117 SURF (allowing to develop and test your application within the comfort of
118 the simulator) and another suited for deployment on real platforms
119 (allowing the resulting application to be highly portable and extremely
122 Even if you do not plan to run your code for real, you may want to switch
123 to GRAS if you intend to use MSG in a very intensive way (e.g. for
124 simulating a peer-to-peer environment).
126 See the \ref GRAS_API section for more details.
128 If your favorite programming environment/model is not there (BSP,
129 components, etc.) is not represented in the SimGrid toolkit yet, you may
130 consider adding it. You should contact us first, though.
132 Any question, remark or suggestion are welcome on the
133 <a href=https://listes.ens-lyon.fr/wws/info/simgrid2-users>SimGrid users
138 \section people People
140 The authors of SimGrid are:
142 - Henri Casanova <casanova#cs.ucsd.edu>
143 - Arnaud Legrand <arnaud.legrand#imag.fr>
144 - Martin Quinson <mquinson#debian.org>
146 \subsection contributers Contributers and alumni project members
148 - Loris Marchal: wrote the new algorithm for simulation TCP
150 - Julien Lerouge : wrote a XML parser for ENV descriptions and helped for
151 the general design during a 4 month period (march-june 2002)
153 - Clément Menier and Marc Perache : wrote a first prototype of the MSG
154 interface during a project at ENS-Lyon (jan 2002).
155 - Dmitrii Zagorodnov : wrote some parts of the first version of SimGrid
160 \section publications Selected publications
162 When citing SimGrid, the prefered reference paper is <i>Scheduling
163 Distributed Applications: the SimGrid Simulation Framework</i>, even if it's
164 a bit old now. We are actively working on improving this.
166 \subsection simulation About simulation
168 \li <b>Scheduling Distributed Applications: the
169 SimGrid Simulation Framework</b>\n
170 by <em>Henri Casanova and Arnaud Legrand and Loris Marchal</em>\n
171 Proceedings of the third IEEE International Symposium
172 on Cluster Computing and the Grid (CCGrid'03)\n
173 Since the advent of distributed computer systems an active field
174 of research has been the investigation of scheduling strategies
175 for parallel applications. The common approach is to employ
176 scheduling heuristics that approximate an optimal
177 schedule. Unfortunately, it is often impossible to obtain
178 analytical results to compare the efficacy of these heuristics.
179 One possibility is to conducts large numbers of back-to-back
180 experiments on real platforms. While this is possible on
181 tightly-coupled platforms, it is infeasible on modern distributed
182 platforms (i.e. Grids) as it is labor-intensive and does not
183 enable repeatable results. The solution is to resort to
184 simulations. Simulations not only enables repeatable results but
185 also make it possible to explore wide ranges of platform and
186 application scenarios.\n
187 In this paper we present the SimGrid framework which enables the
188 simulation of distributed applications in distributed computing
189 environments for the specific purpose of developing and evaluating
190 scheduling algorithms. This paper focuses on SimGrid v2, which
191 greatly improves on the first version of the software with more
192 realistic network models and topologies. SimGrid v2 also enables
193 the simulation of distributed scheduling agents, which has become
194 critical for current scheduling research in large-scale platforms.
195 After describing and validating these features, we present a case
196 study by which we demonstrate the usefulness of SimGrid for
197 conducting scheduling research.
200 \li <b>A Network Model for Simulation of Grid Application</b>\n
201 by <em>Henri Casanova and Loris Marchal</em>\n
203 In this work we investigate network models that can be
204 potentially employed in the simulation of scheduling algorithms for
205 distributed computing applications. We seek to develop a model of TCP
206 communication which is both high-level and realistic. Previous research
207 works show that accurate and global modeling of wide-area networks, such
208 as the Internet, faces a number of challenging issues. However, some
209 global models of fairness and bandwidth-sharing exist, and can be link
210 withthe behavior of TCP. Using both previous results and simulation (with
211 NS), we attempt to understand the macroscopic behavior of
212 TCP communications. We then propose a global model of the network for the
213 Grid platform. We perform partial validation of this model in
214 simulation. The model leads to an algorithm for computing
215 bandwidth-sharing. This algorithm can then be implemented as part of Grid
216 application simulations. We provide such an implementation for the
217 SimGrid simulation toolkit.\n
218 ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2002/RR2002-40.ps.gz
221 \li <b>MetaSimGrid : Towards realistic scheduling simulation of
222 distributed applications</b>\n
223 by <em>Arnaud Legrand and Julien Lerouge</em>\n
224 Most scheduling problems are already hard on homogeneous
225 platforms, they become quite intractable in an heterogeneous
226 framework such as a metacomputing grid. In the best cases, a
227 guaranteed heuristic can be found, but most of the time, it is
228 not possible. Real experiments or simulations are often
229 involved to test or to compare heuristics. However, on a
230 distributed heterogeneous platform, such experiments are
231 technically difficult to drive, because of the genuine
232 instability of the platform. It is almost impossible to
233 guarantee that a platform which is not dedicated to the
234 experiment, will remain exactly the same between two tests,
235 thereby forbidding any meaningful comparison. Simulations are
236 then used to replace real experiments, so as to ensure the
237 reproducibility of measured data. A key issue is the
238 possibility to run the simulations against a realistic
239 environment. The main idea of trace-based simulation is to
240 record the platform parameters today, and to simulate the
241 algorithms tomorrow, against the recorded data: even though it
242 is not the current load of the platform, it is realistic,
243 because it represents a fair summary of what happened
244 previously. A good example of a trace-based simulation tool is
245 SimGrid, a toolkit providing a set of core abstractions and
246 functionalities that can be used to easily build simulators for
247 specific application domains and/or computing environment
248 topologies. Nevertheless, SimGrid lacks a number of convenient
249 features to craft simulations of a distributed application
250 where scheduling decisions are not taken by a single
251 process. Furthermore, modeling a complex platform by hand is
252 fastidious for a few hosts and is almost impossible for a real
253 grid. This report is a survey on simulation for scheduling
254 evaluation purposes and present MetaSimGrid, a simulator built
256 ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2002/RR2002-28.ps.gz
258 \li <b>SimGrid: A Toolkit for the Simulation of Application
260 by <em>Henri Casanova</em>\n
261 Advances in hardware and software technologies have made it
262 possible to deploy parallel applications over increasingly large
263 sets of distributed resources. Consequently, the study of
264 scheduling algorithms for such applications has been an active area
265 of research. Given the nature of most scheduling problems one must
266 resort to simulation to effectively evaluate and compare their
267 efficacy over a wide range of scenarios. It has thus become
268 necessary to simulate those algorithms for increasingly complex
269 distributed, dynamic, heterogeneous environments. In this paper we
270 present SimGrid, a simulation toolkit for the study of scheduling
271 algorithms for distributed application. This paper gives the main
272 concepts and models behind SimGrid, describes its API and
273 highlights current implementation issues. We also give some
274 experimental results and describe work that builds on SimGrid's
276 http://grail.sdsc.edu/papers/simgrid_ccgrid01.ps.gz
278 \subsection research Papers using SimGrid results
280 \li <b>Optimal algorithms for scheduling divisible workloads on
281 heterogeneous systems</b>\n
282 by <em>Olivier Beaumont and Arnaud Legrand and Yves Robert</em>\n
283 In this paper, we discuss several algorithms for scheduling
284 divisible loads on heterogeneous systems. Our main contributions
285 are (i) new optimality results for single-round algorithms and (ii)
286 the design of an asymptotically optimal multi-round algorithm. This
287 multi-round algorithm automatically performs resource selection, a
288 difficult task that was previously left to the user. Because it is
289 periodic, it is simpler to implement, and more robust to changes in
290 the speeds of processors or communication links. On the theoretical
291 side, to the best of our knowledge, this is the first published
292 result assessing the absolute performance of a multi-round
293 algorithm. On the practical side, extensive simulations reveal
294 that our multi-round algorithm outperforms existing solutions on a
295 large variety of platforms, especially when the
296 communication-to-computation ratio is not very high (the difficult
298 ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2002/RR2002-36.ps.gz
299 \li <b>On-line Parallel Tomography</b>\n
300 by <em>Shava Smallen</em>\n
301 Masters Thesis, UCSD, May 2001
302 \li <b>Applying Scheduling and Tuning to On-line Parallel Tomography </b>\n
303 by <em>Shava Smallen, Henri Casanova, Francine Berman</em>\n
304 in Proceedings of Supercomputing 2001
305 \li <b>Heuristics for Scheduling Parameter Sweep applications in
306 Grid environments</b>\n
307 by <em>Henri Casanova, Arnaud Legrand, Dmitrii Zagorodnov and
308 Francine Berman</em>\n
309 in Proceedings of the 9th Heterogeneous Computing workshop
310 (HCW'2000), pp349-363.