Self-adaptive arallel rocessing architecture for high-speed networking

This article describes a packet processing methodology, which executes the required protocol layer functions of a networking device in parallel. Based on the dynamic prediction of the inherent protocol-stack of receiving packets, the data dependency of the layers is speculatively resolved and the functions are processed. Consequently, packet processing latency can be minimized and end-to-end transmission delays can be optimized without sacrificing flexibility in the supported networking protocols and applications. Utilizing the self-adaptive processing methodology, presented in this paper, transmission systems may realize a mean system processing time reduction and consequently a transmission rate increase of up to 40 percent dependent on the quality of the prediction.