Information Technology Reference
In-Depth Information
grid environment where a consumer and a service provider are distributed geographi-
cally across multiple administrative domains. Choosing the suitable resource for a
user job to meet predefined constraints such as deadline, speedup and cost of execu-
tion is an important problem in grids. In our approach, we highly have solved some of
these problems [2]. In general, a Grid broker must make resource selection decisions
in an environment where it has no control over the local resources. The resources are
distributed, and information about the resources is often limited or dated.
Many brokers use the GRIS (Grid Resource Information Service) in discovery and
resource selection phase. The Grid Resource Information Service (GRIS) has been
considered as a tool of inquiring resources on a computational grid for their current
configuration, capabilities, and status. The GRIS is a distributed information service
that can answer queries about a particular resource by directing the query to an infor-
mation provider deployed as part of the Globus services on a grid resource. Using
GRIS permanently cannot be reliable because all data about resource and its perform-
ance are not exist or maybe old. In proposed approach we use a local database to
record any things and events about submitted tasks (job) and at the next time this
information (in local DB) can be very helpful to obtain a background about past re-
source's treatment in order to prediction or another operations. In this paper we will
not do a resource discovery method, but in fact we present a novel way for selecting
the best nodes in pool of discovered nodes. Resource selection involves a set of fac-
tors including application execution time, available main memory, disk (secondary
memory), resource access policies, etc. resource selection must consider information
about resource reliability, prediction error probability, and real time execution. How-
ever, these various performance measures can be considered under the condition that
the middleware allows adaptation of its internal scheduling with desired application's
services. We have considered all of these factors in our approach. Also to reach for
better selection we used the Decision Tree with Fuzzy Logic theory [3]. Induced deci-
sion trees are an extensively-researched solution to classification tasks but general
Decision Tree have some weakness that they covered by Fuzzy Decision Tree (FDT).
The rest of this paper is organized as follows. Section 2 refers to previous research
on resource brokering and scheduling. Section 3 describes Fuzzy Decision Tree Algo-
rithm in our method and also we have described why FDT was used in this paper. In
section 4 we will discusses on proposed architecture and provided applications for
Grid resource broker, respectively. The experimental results and performance evalua-
tion for our method has mentioned in section 5. Finally, section 6 concludes the paper.
2 Related Works
Many projects, such as DI-GRUBER [5], eNANOS [6], AppLes [7] and OGSI- based
broker [4] have been performed on grid. In this section we introduce some of these
brokers.
DI-GRUBER [5], an extension to the GRUBER brokering framework, was devel-
oped as a distributed grid USLA based resource broker that allows multiple decision
points to coexist and cooperate in real-time. GRUBER has been implemented in both
Globus Toolkit4 (GT4) and Globus Toolkit3 (GT3). The part of DI-GRUBER that
dosing resource finding and selecting is called The GRUBER engine. GRUBER
Search WWH ::
Custom Search