A Smart Load Balance Scheme for an Automatic Arbitrage Detection System

Abstract

Models that detect arbitrage opportunities in capital markets require real-time computing intensive systems that exhibit fast system throughput. A key characteristic of such models is that the complexity and consequently the execution time varies from a few milliseconds to several minutes. The response time of such a system is critical since arbitrage opportunities rarely last for more than a few minutes. An HPCN system is an ideal platform for implementing applications that exhibit the above characteristics. Yet, these requirements cannot be easily met unless special consideration is given to the way jobs (arbitrage models) are scheduled and dispatched in the distributed environment where the application is executed. Thus, the need for a self-adapted load balancing mechanism is arising. Such a load balance scheme must be scalable to accommodate for increased workload requirements by making efficient usage of the available resources. Since the application is executed on a multi-user environment consisting of heterogeneous computing elements, (i.e. personal computers, workstations) with varying availability of processing resources, the load balance scheme must ensure the portability and the interoperability of the application