A lockfree shared memory design for high‐throughput multicore packet traffic capture

Abstract

SUMMARYEarlier study by this author has shown that traffic can be captured in multicore architectures with negligible overhead, thus introducing the topic of multicore parallelization into packet traffic analysis. Designs that involve locking or message passing are considered default in parallel processing, which is incompatible with packet traffic with its high throughput and small grain. This paper proposes a new design for shared memory which can be used by any number of processes completely lockfree. Analysis of processing overhead and C/C + + implementation of the design are presented. Copyright © 2014 John Wiley & Sons, Ltd.