A Protocol Offload Framework Applied to a TCP Offload Engine and Web Traffic

A model for protocol offload is used to estimate the CPU utilization before and after TCP/IP protocol extraction. The model presented here is based on queuing theory and the Extensive Message Oriented (EMO) model. A modified version of Apache 2.2 Web server with raw sockets and a PC acting as the offload engine are used to implement the offload engine. Benchmarks will be run on our implementation to estimate the impact of protocol offload extraction and to refine our analytical model. The utilization of a non-offload system is the aggregation of the time the server is serving the (1) network application, (2) the protocols, and, (3) any other non-network application. Therefore: