Mellanox ASAP2
定义
Accelerated Switching and Packet Processing
加速交换与数据报处理技术,原由Mellanox提出但在Nvidia收购后更名为 Nvidia ASAP2
应对场景
主要针对高速网络下的加速网络数据包场景,传统网络数据包处理需要将网络通信数据包全部提交至CPU进行处理,这些处理过程占用了CPU资源并且开销日益增大
应对思路
让网卡接管网络IO的处理,并且在硬件上完成原本CPU需要进行的网络数据解析任务,并且携带网络数据分流的特性,支持Single Root Virtualization等方式模拟出多个网络设备,将不同的网络流量分别呈递,能够接入[[OpenVirtualSwitch]] 虚拟交换或其它虚拟网络,使得这一过程对CPU的开销降低到非常低的程度,从而达到硬件加速的效果
性能
1 | Benchmarks show that on a server with a 25G interface, OVS accelerated by ASAP2 Direct achieves 33 million packets per second (mpps) for a single flow with near-zero CPU cores consumed by the OVS data plane, and about 18 mpps with 60,000 flows performing VXLAN encap/decap. These performance results are three to ten times higher than DPDK-accelerated OVS. ASAP2 offers the best of both worlds, software-defined flexible network programmability, and high network I/O performance for the state-of-art speeds from 10G to 25/40/50/100/200G. |
25G 端口下,通过ASAP2 Direct
技术能够处理单条流33mpps数据包,且无CPU占用,在VXLAN
解包生成包操作下能够达到最多6万条流一共18mpps的性能
目前此技术可应用在25/40/50/100/200G网络通信模式中