Access control based on performance estimation and client clustering in large-scale distributed systems

A large-scale distributed system provides an attractive scalable infrastructure for network services. However, the loosely coupled nature of this environment will make data access unpredictable. Today, an increasing number of network applications require not only considerations of server computation power but also accessibility for adequate job allocations. An effective mechanism of access control is important in this environment. In this paper, we present a mechanism of access control based on performance estimation and clients clustering called ACPC that can reduce the data access cost and guarantee load balance and adaptivity. With the experimental results, we show that ACPC performs better than other access control algorithms.