Information Technology Reference
In-Depth Information
the efi ciency, performance, and reliability of new algorithms on large
topologies before considering their implementation on test-beds and pro-
duction systems.
In particular, the study of grids is signii cantly aided by robust and
rapid prototyping via simulation, due to the sheer scale and complexities
that arise when operating over many administrative domains, which pre-
cludes easy prototyping on real test-beds. Grid computing [1] has been
integral in enabling knowledge breakthroughs in i elds as diverse as cli-
mate modeling, drug design, and protein analysis, through the harness-
ing of computing, network, sensor, and storage resources owned and
administered by many different organizations. These i elds (and other so-
called “grand challenges”) have benei ted from the economies of scale
brought about by grid computing, tackling difi cult problems that would
be impossible to feasibly solve using the computing resources of a single
organization. However, when prototyping such applications and services
that harness the power of the grid, it is benei cial to test their operation via
simulation in order to optimize their behavior, and avoid placing strain on
grid resources during the development phase.
Despite the obvious advantages of simulation when prototyping appli-
cations and services that run on grids, realistically simulating large topol-
ogies and complicated scenarios can take signii cant amounts of memory
and computational power. For statistical signii cance, large numbers of
simulation runs are needed to increase our coni dence in the results we
obtain from simulation platforms. This is particularly the case when
studying applications and services that store and move signii cant vol-
umes of data over the grid, such as data-grids or content and service deliv-
ery networks. Simulators that attempt to model the full complexity of
TCP/IP networking in such environments scale poorly and often run sig-
nii cantly slower than real time, practically defeating the purpose of simu-
lating such environments in the i rst place.
In this chapter we look at incorporating l ow-level (or “l uid”) network-
ing models into grid simulators, in order to improve the scalability and
speed of grid simulations by reducing the overhead of data- and network-
intensive experiments, and improving their accuracy. Network l ow mod-
els are used that closely approximate actual steady-state TCP/IP network-
ing. We utilize the GridSim toolkit as a candidate implementation, and
fully replace the existing packet-level networking model in GridSim with
a l ow-level networking stack. However, the principles outlined in this
chapter could be applied to other simulation platforms.
The remainder chapter is organized as follows. The next section describes
the GridSim Toolkit and gives a brief overview of its feature set. The exist-
ing packet-level networking implementation for the GridSim toolkit is then
described, and some inefi ciencies are identii ed that arise when doing
large-scale network and data centric simulations. We then outline the basic
Search WWH ::
Custom Search