Publication Date

2006

Degree Type

Master's Project

Degree Name

Master of Science (MS)

Department

Computer Science

Abstract

In this project, we design a cooperative interval caching (CIC) algorithm for clustered video servers, and evaluate its performance through simulation. The CIC algorithm describes how distributed caches in the cluster cooperate to serve a given request. With CIC, a clustered server can accommodate twice (95%) more number of cached streams than the clustered server without cache cooperation. There are two major processes of CIC to find available cache space for a given request in the cluster: to find the server containing the information about the preceding request of the given request; and to find another server which may have available cache space if the current server turns out not to have enough cache space. The performance study shows that it is better to direct the requests of the same movie to the same server so that a request can always find the information of its preceding request from the same server. The CIC algorithm uses scoreboard mechanism to achieve this goal. The performance results also show that when the current server fails to find cache space for a given request, randomly selecting a server works well to find the next server which may have available cache space. The combination of scoreboard and random selection to find the preceding request information and the next available server outperforms other combinations of different approaches by 86%. With CIC, the cooperative distributed caches can support as many cached streams as one integrated cache does. In some cases, the cooperative distributed caches accommodate more number of cached streams than one integrated cache would do. The CIC algorithm makes every server in the cluster perform identical tasks to eliminate any single point of failure, there by increasing availability of the server cluster. The CIC algorithm also specifies how to smoothly add or remove a server to or from the cluster to provide the server with scalability.

Download

Share

COinS