Implementing Global Buffer Management for the Global File System

With the advance of high-speed network technology, such as ATM, Gigabit Ethernet, and Fiber Channel, the degree of sharing computing resources (i.e., processor and memory) in a distributed environment becomes more and more expanded. It is very important to reduce the number of disk I/O by efficiently utilizing a global memory in a distributed shared file system because there exists performance gaps between remote memory access and disk access. In this paper, we describe the design and implementation of global memory management for a large shared storage system. In our implementation, global buffer manager is simplified to reduce communication overheads and eliminate bottleneck problem. We examined performance trade-offs by using simulation study while varying several performance parameters