Abstract:
Large scale distributed systems such as computational and data grids require pervasive
end-to-end resource monitoring for effective problem diagnoses, performance analysis,
performance tuning and job scheduling. Gathering monitoring information can be a
tedious task in itself. The complexity in retrieving such an information originates from the
fact that not only we require knowledge of end systems but also the intervening network
segments, over which we may not have any control or administrative privileges, like the
Internet. The End host monitoring Agent (EMA) is a java based, dynamically
configurable monitoring tool, which enables the user for easy data acquisition from endhosts.
The information comprises of static and dynamic performance data such as the
system, memory, CPU, disk and network parameters. It has built-in mechanisms for
regularly scheduled measurements and reporting of results to any data repository. EMA
is a lightweight tool and has a scalable, multilayered architecture. In this report we have
discussed the need for such an application, its architecture, design and applications in
detail. In the end results and some future tasks are enlisted.
