Multi-Table Programmable Parser with online flow-level update consistency for satellite networks

Abstract

Programmable packet parsers in the space data plane, responsible for identifying protocols and extracting keywords on demand, are significant to realize protocol upgrading for satellite networks. Because flow-level update consistency is significant, onboard resources are limited, and space radiation may cause errors in packet parsers, programmable packet parsers with online flow-level update consistency, low complexity, and high reliability are urgently needed. In our previous work, a Multi-Table Programmable Parser (MTPP) [1] is designed for fast configuration and low storage cost by networking configuration and multiple tables. In this paper, we propose a Multi-Table Programmable Parser with Online Flow-Level Update Consistency (MTPP-OFLUC), consisting of five tables and eight models. To guarantee the flow-level update consistency, each table consists of a main table and a backup table. Moreover, to improve the reliability of MTPP-OFLUC, a fault tolerance mechanism is designed. Furthermore, an Online Configuration based on Super Parse Graph Segmentation (OC-SPGS) is proposed to achieve fast configuration with flow-level update consistency. The numerical analysis under typical parse graphs demonstrates that MTPP-OFLUC uses less storage than that of PISA-based parser. Due to backup tables adapted to guarantee the flow-level update consistency, the storage cost of MTPP-OFLUC is only 11.53% more than that of MTPP. Through OC-SPGS, the parsing success ratio of MTPP-OFLUC is up to 66.7% larger than that of MTPP. Finally, a prototype of MTPP-OFLUC is implemented on Xilinx Virtex-7 FPGA. The response time of MTPP-OFLUC and the network is 1.460 us and 4.196 ms, respectively. The size of configuration information is less than that of PISA-based parser. In summary, rapid response and flow-level update consistency are realized by MTPP-OFLUC. Moreover, security risks, user-friendly interfaces, potential capacity of upgrades, and supports for higher data rates still need to be considered in the future works.