An adaptive model for building service-partition system

Many software clustering approaches have been proposed for developing enterprise systems in the distributed environment to gain improved performance, good scalability and high availability. The mostly adopted approaches — symmetrical and unsymmetrical clustering, have some negative impact on system performance. Service-partition is a new clustering approach which solves these disadvantages. It divides the core business service into partitions based on data and distributes them on each node hence it enables the “aggressive cache” and reduces the remote invocations and database contention. This paper presents a practical model named Adaptive Partition Model (APM) for building enterprise systems using the service-partition approach. APM consists of the business level and the meta-level which provides an adaptive characteristic that can balance the workload dynamically. A case study is also reported in reengineering a standalone financial trading system using APM. The test results show the reengineered trading system has improved performance, scalability and workload distribution than before.