![](bg1.png)
Simulation of Real-time Data Grid Systems via DGridSim Simulator
Safai Tandoğan
1
, Mustafa Müjdat Atanak
2
and Atakan Doğan
3
1
C. Tech, TUBITAK MAM TEKSEB, Kocaeli, Turkey
2,3
Department of Electrical and Electronics Engineering, Anadolu University, 26470 Eskisehir, Turkey
Keywords: Data Grid, Real-time, Job Scheduling, Data Dissemination, Data Replication.
Abstract: In this study, DGridSim simulator will be introduced and some example simulation results will be
presented. DGridSim can simulate four different Data Grid system organizations. Furthermore, for every
system organization, the simulation of job scheduling, data dissemination, and data replication algorithms
are supported, while all related system resources including computing, data storage, and network are
reserved in advance in order to meet deadlines associated with jobs. DGridSim simulator is designed to be
modular and easily extensible.
1 INTRODUCTION
Data Grid systems are highly distributed systems
that are increasingly used in the analysis of large
amounts of data (Chervenak et al., 2000). There
exist a large number of parameters that affect the
performance of these sophisticated systems. Some of
these parameters change dynamically. Hence,
developing analytical models to study the impact of
any system parameter is a nontrivial task and, often,
simulators are used. Hence, a number of simulators
exist in the literature (Bell et al., 2003, Buyya et al.
2002, Lamehamedi et al. 2003, Casanova et al.
2003).
Any real-time Data Grid system should have a
set of services to support job scheduling, data
replication, data dissemination, and advance
reservation. Based on this rationale, this study
proposes a unique framework, DGridSim, for
simulating four different real-time Data Grid
systems based on well-defined services.
2 DGRIDSIM MODELS
DGridSim supports four different Data Grid system
models, some of which are based on the studies in
the literature, and the others are proposed herein.
Furthermore, DGridSim provides a unified platform
for simulating job scheduling, data dissemination,
and data replication algorithms for all four models.
The Data Grid system models supported by
DGridSim are listed in Table 1, whose details are
provided in (DGridSim Project).
Table 1: Data Grid system models of DGridSim.
Model I: In order to simulate Data Grid systems
of Model I, Model I should be chosen under Grid
Model segment in General tab of the GUI. Figure 1
shows a sample screenshot of the GUI when Model I
is chosen.
In Model I, job scheduling is carried out in
hierarchical fashion. In DGridSim, seven different
Grid scheduling algorithms are realized: Random,
EDF (Earliest Deadline First), MCTF (Minimum
Completion Time First), MCTFwDP (Minimum
Completion Time First with Data Present), MMwDP
(MinMin with Data Present), RT MCTFwDS (Real-
Time Minimum Completion Time First with Data
Staging) and RT MMwDP (Real-Time MinMin with
Data Present). DGridSim is equipped with a sample
site scheduling algorithm, namely RT Max Max
(Real-Time MaxMax).
Data dissemination is carried out in hierarchical
fashion as well. DGridSim currently supports only
Minimum Delay Feasible Path First algorithm as a
part of its Data Management Service.
Model I implementation of DGridSim offers four
182
Tando
˘
gan S., Atanak M. and Do
˘
gan A..
Simulation of Real-time Data Grid Systems via DGridSim Simulator.
DOI: 10.5220/0004057801820185
In Proceedings of the 2nd International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2012),
pages 182-185
ISBN: 978-989-8565-20-4
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)