eModel: Addressing the Need for a Flexible Modeling Framework in Autonomic Computing

This paper describes a novel, exible framework, eModel , designed to address the run-
time requirements of autonomic computing: on-line workload measurement, analysis,
and prediction. The eModel architecture has been developed using platform indepen-
dent technology (XML and Java) to allow for maximum portability while also allowing
for ease-of-integration with existing measurement and system management tools. The
eModel toolkit consists of a GUI based model builder tool, a data base deployment
tool, a runtime tool, and an analysis tool. In addition to the toolkit, the eModel
design provides a runtime architecture which can be deployed directly without using
any interaction with the GUI. The architecture is exible enough to allow for incorpo-
ration with models of various complexity, including modeling techniques that require
a hierarchical approach to attain reasonable accuracy. We present examples that il-
lustrate eModel as a capacity planning tool as well as an augmentation to autonomic
system management in an effort to highlight the technological gaps that the eModel
framework is capable of bridging.