Adaptive workflow policy-based service management scheme in physio-grid

Abstract

Physio-Grid is a new type of heart disease detection system suggested by Changhee Han. Existing diagnosis service does not provide integrated service model for heart disease detection to doctor. By using grid technology, Phyio-Grid provides the service platform that integrates computing capability and medical data across hospitals. However, this Physio-Grid does not consider the service level between service user and service provider. In medical application, it can be serious problem because it is related with life. To solve that, in this thesis, we discuss service level management in Physio-Grid and identify three service level (Premium, Standard, and Lite) considering user``s service requirement and propose the scheme for managing the different service level as well as minimizing the resource operation cost. The scheme describes each service level using workflow that is implemented with Kepler. For execution, we used the PQRM Plus which is policy-based grid resource management system. Especially, we proposed the adaptive workflow policy-management scheme to control PQRM Plus integrated Kepler workflow. The idea for policy management is to use estimated computing resources which is derived with upper bound and lower bound. Comparing computing resource model used in discrete event system and real grid resource throughput, we generate optimized policies that satisfy service level and minimize resource cost. We evaluated this scheme to the Physio-Grid service and show our scheme could guarantee the requirement of each service level as well as minimize the resource cost.