Abstract
Distributed object-oriented platforms have become important components for parallel and distributed computing and service frameworks. Among distributed object-oriented software, Microsoft .NET Remoting is one of the key methods for performing parallel and distributed computing in .NET environments. In this thesis, we present our methodologies in supporting .NET Remoting over cluster environments with a load balancing mechanism. We take advantage of the programmability of Intel IXP1200 network processor to develop a load balancing getaway that implementing packet classification instead of protocol demultiplexing. IXP1200 contains a StrongARM of 232 MHz and six programmable 32-bit RISC processors of 232 MHz (a.k.a. Microengine). We divid the load balancer into two parts, one is the control system executed in StrongARM and the other one is data path system executed in Microengines. Stateful supports for Remoting services are also incorporated. The system will be conscious of the session created between client and server and dispatch jobs by the session table. Experiments show that our scheme can significantly enhance the system throughput (up to 55%) and reduce the response time (up to 56%) compared with Microsoft Network Load Balancing technology when the traffic is heavy. Besides, we also discussed the effect of Microengine allocation and found out which part of system consumes the most computing power.