LAN file server performance: a processor sharing model

Considers a system of single user workstations connected to a file server over a local network. The workstations request file transfers from the server in a random manner. In response to the first request, a major portion of the file is transferred from the server to the workstation local memory, while response to subsequent requests is in smaller chunks. In the paper, the authors investigate the benefits of employing the round-robin/processor-sharing policy, for allocating the server capacity between the two types of requests. They generate a Markovian state description of the system and analyze it using a continuous parameter Markovian chain.