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