GASPP: A GPU-Accelerated Stateful Packet Processing Framework

Graphics processing units (GPUs) are a powerful plat-form for building high-speed network traffic processing applications using low-cost hardware. Existing systems tap the massively parallel architecture of GPUs to speed up certain computationally intensive tasks, such as cryp-tographic operations and pattern matching. However, they still suffer from significant overheads due to critical-path operations that are still being carried out on the CPU, and redundant inter-device data transfers. In this paper we present GASPP, a programmable net-work traffic processing framework tailored to modern graphics processors. GASPP integrates optimized GPU-based implementations of a broad range of operations commonly used in network traffic processing applica-tions, including the first purely GPU-based implementa-tion of network flow tracking and TCP stream reassem-bly. GASPP also employs novel mechanisms for tackling control flow irregularities across SIMT threads, and shar-ing memory context between the network interface and the GPU. Our evaluation shows that GASPP can achieve multi-gigabit traffic forwarding rates even for computa-tionally intensive and complex network operations such as stateful traffic classification, intrusion detection, and packet encryption. Especially when consolidating mul-tiple network applications on the same device, GASPP achieves up to 16.2 × speedup compared to standalone GPU-based implementations of the same applications.