5 <img align=center src="simgrid_logo.png" alt="SimGrid"><br>
9 \section overview Overview
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 \section people People
20 The authors of SimGrid are:
22 \author Henri Casanova <casanova@cs.ucsd.edu>
23 \author Arnaud Legrand <arnaud.legrand@imag.fr>
24 \author Martin Quinson <martin.quinson@tuxfamily.org>
26 \section intro Available Softwares
28 The SimGrid toolkit is composed of different modules :
30 \li XBT (eXtensive Bundle of Tools) is a portable library with many
31 convenient portable datastructures (vectors, hashtables, heap,
32 contexts ...). Most other SimGrid modules rely on it.
34 \li SURF provides the core functionnalities to simulate a virtual
35 platform. It is very low-level and is not intended to be used as
36 such but rather to serve as a basis for higher-level simulators
37 (like MSG, GRAS, SMPI, ...). It relies on a fast max min linear
40 \li MSG is a simulator built using the previous modules. It aims at
41 being realistic and is application-oriented. It is the software layer
42 of choice for building simulation with multiple scheduling agents.
44 \li GRAS (<em>not functionnal yet</em>) is an ongoing project to emulate virtual virtual platforms
45 through SURF. As a consequence a code developped using the GRAS
46 framework is able to run as well in the real-world as in the
47 simulator. The resulting code is very portable and highly interoperable while
48 remaining very efficient. Even if you do not plan to run your code for real,
49 you may want to switch to GRAS if you intend to use MSG in a very intensive way
50 (e.g. for simulating a peer-to-peer environment).
52 \li SMPI (<em>not functionnal yet</em>) is an ongoing project to enable MPI code to run on top of a
53 virtual platform through SURF. It follows the same principle as
54 the ones used in GRAS but is specific to MPI applications.
56 The section \ref publications contains links to papers that provide
57 additional details on the project as well as validation and
60 The software can be downloaded from <a href="http://gcl.ucsd.edu/SimGrid/dl/">here</a>.
62 \section install Installation
65 \li <tt>./configure</tt>
66 \li <tt>make all install</tt>
68 If you are not familiar with compiling C files under UNIX and using
69 libraries, you will find some more informations in Section \ref
72 \section documentation API Documentation
74 The API of all different modules is described in \ref SimGrid_API.
76 See \ref SimGrid_examples for an introduction on the way to use these modules.
78 \section users_contributers Users / Contributers
80 \subsection contributers Contributers
82 \li Loris Marchal: wrote the new algorithm for simulation TCP
84 \li Julien Lerouge : wrote a XML parser for ENV descriptions and
85 helped for the general design during a 4 month period (march-june 2002)
87 \li Clément Menier and Marc Perache : wrote a first prototype of
88 the MSG interface during a project at ENS-Lyon (jan 2002).
89 \li Dmitrii Zagorodnov : wrote some parts of the first version
92 \subsection mailinglist User Mailing List
94 We have a <a href=https://listes.ens-lyon.fr/wws/info/simgrid2-users> mailing list for
97 \section publications Publications
99 \subsection simulation About simulation
101 \li <b>Scheduling Distributed Applications: the
102 SimGrid Simulation Framework</b>\n
103 by <em>Henri Casanova and Arnaud Legrand and Loris Marchal</em>\n
104 Proceedings of the third IEEE International Symposium
105 on Cluster Computing and the Grid (CCGrid'03)\n
106 Since the advent of distributed computer systems an active field
107 of research has been the investigation of scheduling strategies
108 for parallel applications. The common approach is to employ
109 scheduling heuristics that approximate an optimal
110 schedule. Unfortunately, it is often impossible to obtain
111 analytical results to compare the efficacy of these heuristics.
112 One possibility is to conducts large numbers of back-to-back
113 experiments on real platforms. While this is possible on
114 tightly-coupled platforms, it is infeasible on modern distributed
115 platforms (i.e. Grids) as it is labor-intensive and does not
116 enable repeatable results. The solution is to resort to
117 simulations. Simulations not only enables repeatable results but
118 also make it possible to explore wide ranges of platform and
119 application scenarios.\n
120 In this paper we present the SimGrid framework which enables the
121 simulation of distributed applications in distributed computing
122 environments for the specific purpose of developing and evaluating
123 scheduling algorithms. This paper focuses on SimGrid v2, which
124 greatly improves on the first version of the software with more
125 realistic network models and topologies. SimGrid v2 also enables
126 the simulation of distributed scheduling agents, which has become
127 critical for current scheduling research in large-scale platforms.
128 After describing and validating these features, we present a case
129 study by which we demonstrate the usefulness of SimGrid for
130 conducting scheduling research.
133 \li <b>A Network Model for Simulation of Grid Application</b>\n
134 by <em>Henri Casanova and Loris Marchal</em>\n
136 In this work we investigate network models that can be
137 potentially employed in the simulation of scheduling algorithms for
138 distributed computing applications. We seek to develop a model of TCP
139 communication which is both high-level and realistic. Previous research
140 works show that accurate and global modeling of wide-area networks, such
141 as the Internet, faces a number of challenging issues. However, some
142 global models of fairness and bandwidth-sharing exist, and can be link
143 withthe behavior of TCP. Using both previous results and simulation (with
144 NS), we attempt to understand the macroscopic behavior of
145 TCP communications. We then propose a global model of the network for the
146 Grid platform. We perform partial validation of this model in
147 simulation. The model leads to an algorithm for computing
148 bandwidth-sharing. This algorithm can then be implemented as part of Grid
149 application simulations. We provide such an implementation for the
150 SimGrid simulation toolkit.\n
151 ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2002/RR2002-40.ps.gz
154 \li <b>MetaSimGrid : Towards realistic scheduling simulation of
155 distributed applications</b>\n
156 by <em>Arnaud Legrand and Julien Lerouge</em>\n
157 Most scheduling problems are already hard on homogeneous
158 platforms, they become quite intractable in an heterogeneous
159 framework such as a metacomputing grid. In the best cases, a
160 guaranteed heuristic can be found, but most of the time, it is
161 not possible. Real experiments or simulations are often
162 involved to test or to compare heuristics. However, on a
163 distributed heterogeneous platform, such experiments are
164 technically difficult to drive, because of the genuine
165 instability of the platform. It is almost impossible to
166 guarantee that a platform which is not dedicated to the
167 experiment, will remain exactly the same between two tests,
168 thereby forbidding any meaningful comparison. Simulations are
169 then used to replace real experiments, so as to ensure the
170 reproducibility of measured data. A key issue is the
171 possibility to run the simulations against a realistic
172 environment. The main idea of trace-based simulation is to
173 record the platform parameters today, and to simulate the
174 algorithms tomorrow, against the recorded data: even though it
175 is not the current load of the platform, it is realistic,
176 because it represents a fair summary of what happened
177 previously. A good example of a trace-based simulation tool is
178 SimGrid, a toolkit providing a set of core abstractions and
179 functionalities that can be used to easily build simulators for
180 specific application domains and/or computing environment
181 topologies. Nevertheless, SimGrid lacks a number of convenient
182 features to craft simulations of a distributed application
183 where scheduling decisions are not taken by a single
184 process. Furthermore, modeling a complex platform by hand is
185 fastidious for a few hosts and is almost impossible for a real
186 grid. This report is a survey on simulation for scheduling
187 evaluation purposes and present MetaSimGrid, a simulator built
189 ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2002/RR2002-28.ps.gz
191 \li <b>SimGrid: A Toolkit for the Simulation of Application
193 by <em>Henri Casanova</em>\n
194 Advances in hardware and software technologies have made it
195 possible to deploy parallel applications over increasingly large
196 sets of distributed resources. Consequently, the study of
197 scheduling algorithms for such applications has been an active area
198 of research. Given the nature of most scheduling problems one must
199 resort to simulation to effectively evaluate and compare their
200 efficacy over a wide range of scenarios. It has thus become
201 necessary to simulate those algorithms for increasingly complex
202 distributed, dynamic, heterogeneous environments. In this paper we
203 present SimGrid, a simulation toolkit for the study of scheduling
204 algorithms for distributed application. This paper gives the main
205 concepts and models behind SimGrid, describes its API and
206 highlights current implementation issues. We also give some
207 experimental results and describe work that builds on SimGrid's
209 http://grail.sdsc.edu/papers/simgrid_ccgrid01.ps.gz
211 \subsection research Papers using SimGrid results
213 \li <b>Optimal algorithms for scheduling divisible workloads on
214 heterogeneous systems</b>\n
215 by <em>Olivier Beaumont and Arnaud Legrand and Yves Robert</em>\n
216 In this paper, we discuss several algorithms for scheduling
217 divisible loads on heterogeneous systems. Our main contributions
218 are (i) new optimality results for single-round algorithms and (ii)
219 the design of an asymptotically optimal multi-round algorithm. This
220 multi-round algorithm automatically performs resource selection, a
221 difficult task that was previously left to the user. Because it is
222 periodic, it is simpler to implement, and more robust to changes in
223 the speeds of processors or communication links. On the theoretical
224 side, to the best of our knowledge, this is the first published
225 result assessing the absolute performance of a multi-round
226 algorithm. On the practical side, extensive simulations reveal
227 that our multi-round algorithm outperforms existing solutions on a
228 large variety of platforms, especially when the
229 communication-to-computation ratio is not very high (the difficult
231 ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2002/RR2002-36.ps.gz
232 \li <b>On-line Parallel Tomography</b>\n
233 by <em>Shava Smallen</em>\n
234 Masters Thesis, UCSD, May 2001
235 \li <b>Applying Scheduling and Tuning to On-line Parallel Tomography </b>\n
236 by <em>Shava Smallen, Henri Casanova, Francine Berman</em>\n
237 in Proceedings of Supercomputing 2001
238 \li <b>Heuristics for Scheduling Parameter Sweep applications in
239 Grid environments</b>\n
240 by <em>Henri Casanova, Arnaud Legrand, Dmitrii Zagorodnov and
241 Francine Berman</em>\n
242 in Proceedings of the 9th Heterogeneous Computing workshop
243 (HCW'2000), pp349-363.