Master Thesis - 2008

Network services based on distributed components are attractive for all service providers, since they can offer high-quality services with lower cost. Components are replicated and placed to many network nodes for load-balancing, and they can be used as a substitute when a components has failed.

In this paper, I proposed a new platform, which is an effective and fault-tolerant platform for such network services. Proposed platform consists of three functions; initial components placement, load-sharing at steady-state phase, failure-recovery when a failure occurs. In this paper, two functions that are initial components placement and load-sharing are proposed. Initial placement function places initial components that are needed for the platform to work and constructs overlay networks among the same functional components. With load-sharing function, each components can autonomously increase and decrease the replicated components and that such replicated components are placed on suitable nodes for load-balancing. Nodes having high performance have capacity to store many components and have latitude to share own performance with storing components. By placing more replicated component on the nodes having high performance, there are few components that have too low processing power and all services are provided at high speed.

Finally, simulation experiment shows that the proposed platform is effective and fault-tolerant for the network services based on distributed components.