Instruction set architectural guidelines for embedded packet-processing engines
In this paper, we present a new architectural concept for network processors called FlexPath NP. The central idea behind FlexPath NP is to systematically map network processor (NP) application sub-functions onto both SW programmable processor (CPU) resources and (re-)configurable HW building blocks, such that different packet flows are forwarded via different, optimized processing paths through the NP. Packets with well understood, relatively simple processing requirements may even bypass the central CPU complex (AutoRoute). In consequence, CPU processing resources are more effectively used and the overall NP performance and throughput are improved compared to conventional NP architectures. We present analytical performance estimations to quantify the performance advantage of FlexPath (expressed as available CPU instructions for each packet traversing the CPUs) and introduce a platform-based System on Programmable Chip (SoPC) based architecture which implements the FlexPath NP concept.
Unfortunately, ACM prohibits us from displaying non-influential references for this paper.
To see the full reference list, please visit http://dl.acm.org/citation.cfm?id=1084904.