Information Technology Reference
In-Depth Information
stored on the network. These hash keys are then used to search for data,
and at every junction of the tree a routing decision is made using the keys
prei x compared with the junctions at the binary branches. However,
P-Grid requires that extra processing for the data be moved to nodes.
Another side-effect in using a binary tree is that a search must always
transverse the tree's i xed paths producing slow and frequent trafi c.
To emphasize, the large scale coupled with the dynamic nature of a grid
environment poses many challenges for information services. In this
chapter, we present PIndex, a grouped P2P model for scalable grid infor-
mation services. PIndex aims to resolve the issues that arise within a
highly dynamic grid environment with its salient features listed below:
It introduces the concept of peer groups so that a large number of
peer nodes in a grid environment can be aggregated into groups
to speed up the lookup process in resource discovery.
It builds on the dynamic aggregate framework of Globus MDS4 to
handle dynamic node presence and data.
It enhances fault resilience through the replication of state infor-
mation among nodes within a peer group.
It reduces the effects of churn, limiting them locally to a peer
group level and isolating the effect to the rest of the network.
We build models using colored Petri nets (CPNs) to evaluate PIndex, and
the simulation results show that PIndex is scalable with a large number of
peer nodes.
The remainder of this chapter is organized as follows. The next section
gives an overview of Globus MDS4. This is followed by an introduction to
PIndex, and an explanation of its main concepts. We then describe the
design and implementation of the CPN models for PIndex evaluations.
Finally, we present the simulation results demonstrating the effectiveness
of PIndex for scalable grid information services.
13.2
Owing to its high dynamism, a grid environment can produce large
amounts of state information that needs to be handled promptly. The
Globus MDS4 is a suite of Web Services Resource Framework (WSRF) [9]
compliant services that can be used to monitor and discover resources
through notii cations. MDS4 enables one to advertise large amounts of
state information to users within a VO, and the ability to monitor and
query these resources. This section gives an overview on MDS4, along
with an extrapolation of its performance when placed in a highly dynamic
grid environment.
An Overview of Globus MDS4
 
 
Search WWH ::




Custom Search