A Processing Element and Programming Methodology for Click Elements

Programmable platform architectures are a compelling alternative to ASICs for network processing applications, but they are notoriously difficult to program. The reason for this is the large discrepancy between the concurrency requirements of applications and the concurrency capabilities of the platforms. Domain-specific languages such as Click are a good way to capture application concurrency, but there are still numerous other opportunities for network processors to provide features that are a closer match for application requirements. In this paper, we propose a processing element that is highly effective at implementing Click application tasks. We provide a productive methodology for programming the PE using Click as a high-level input language. On an FPGA, our PEs can handle over 8 Gbps of traffic for an IPv4 forwarding application.