PacketShader

Abstract

We present PacketShader, a high-performance software router framework for general packet processing with Graphics Processing Unit (GPU) acceleration. PacketShader exploits the massively-parallel processing power of GPU to address the CPU bottleneck in current software routers. Combined with our high-performance packet I/O engine, PacketShader outperforms existing software routers by more than a factor of four, forwarding 64B IPv4 packets at 39 Gbps on a single commodity PC. We have implemented IPv4 and IPv6 forwarding, OpenFlow switching, and IPsec tunneling to demonstrate the flexibility and performance advantage of PacketShader. The evaluation results show that GPU brings significantly higher throughput over the CPU-only implementation, confirming the effectiveness of GPU for computation and memory-intensive operations in packet processing.

SNF: synthesizing high performance NFV service chains

“…Among the several candidate platforms that allow such a representation, we developed our prototype atop Click because it is the most widely used NFV platform in the academia. Many earlier efforts built upon it to improve its performance and scalability, hence we believe that this choice will maximize SNF's impact as it allows direct comparison with state of the art Click variants such as RouteBricks (Dobrescu et al, 2009), PacketShader (Han et al, 2010), Double-Click (Kim et al, 2012), SNAP (Sun & Ricci, 2013), ClickOS (Martins et al, 2014), and FastClick (Barbette, Soldani & Mathy, 2015).…”

“…This first use case targets a direct comparison with the state of the art. Specifically, we chain a popular implementation of a software-based router that, after several years of successful research contributions (Dobrescu et al, 2009;Han et al, 2010;Kim et al, 2012;Sun & Ricci, 2013;Martins et al, 2014;Barbette, Soldani & Mathy, 2015), achieves scalable performance at tens of Gbps.…”

“…Among the first challenges in NFV was to scale software-based packet processing by exploiting the characteristics of modern hardware architectures. To do so, several works leveraged parallelism first across multiple servers and then across multiple cores, sockets, memory controllers, and graphical processing units (GPUs) (Han et al, 2010;Kim et al, 2015b) within a single server (Dobrescu et al, 2009;Dobrescu et al, 2010).…”

Data protection: Combining fragmentation, encryption, and dispersion

“…AGPU acceleration for DCT 8×8 (DCT by block 8×8) is already shown in [Patel…09] or [Obukhov…08], however, not in the context of SE. Another reason for using GPU is their rapid density improvement which goes faster than usual CPU improvement [Han…10].…”

Supporting Low-Latency CPS Using GPUs and Direct I/O Schemes

“…Certain tasks such as software routing, and network encryption can achieve more than an order of magnitude performance gain [2], [4]. Distributed storage systems that employ hash based primitives enjoy a speed up of 8x [1].…”