Evaluation of Forwarding Efficiency in NFV-Nodes Toward Predictable Service Chain Performance

Posted 2021-01-08 grglym

文章名称：Evaluation of Forwarding Efficiency in NFV-Nodes Toward Predictable Service Chain Performance

发表时间：2017

期刊来源：IEEE Transactions on Network and Service Management

解决问题：

所做贡献：本文以实验评估为主，

1、三种数据包I / O架构（NAPI，netmap和DPDK）

2、三种虚拟I / O机制（vhost-net，vhost-user和SR-IOV）

3、四种实用转发程序（Open vSwitch，OVS-DPDK ，xDPd-DPDK和Lagopus）及三个参考程序（Linux Bridge，VALE和L2FWD-DPDK）

不足之处：

实验对比：实验通过对比不同的I/O架构、转发程序、I/O机制。

对比指标：Througput、Output Rate、Latency、InputRate--Output Rate

The concept of network functions virtualization (NFV) has been embodied in commercial networks over the past years. Software-based virtual network functions have forwarding performance concerns in general, and various acceleration technologies have been developed so far, such as DPDK and vhost-user. Existence of several alternatives requires network engineers or operators to select appropriate technologies; however, no pragmatic criterion exists for constructing high-performanceNFV-nodes. From their points of view, a lack of common benchmark and understanding ofperformance characteristics makes it difficult to predict hop-by-hop performance in a service chain, which results in prevention of NFV deployment in mission-critical networks. In this paper, we clarify performance characteristics of packet forwarding in NFV nodes focusing on three types ofacceleration technologies; packet I/O architecture, virtual network I/O, and forwarding engine in a practical stage. We examined three packet I/O architectures (NAPI, netmap, and DPDK), three virtual I/O mechanisms (vhost-net, vhost-user, and SR-IOV), and four practical forwarding programs (Open vSwitch, OVS-DPDK, xDPd-DPDK, and Lagopus) with three referential programs (Linux Bridge, VALE, and L2FWD-DPDK). The experiment was conducted on a 40 GbE environment and we examined two device-under-test machines having different CPU performance. We argue performance characteristics of each technology and give quantitative analyses of the result. The key findings are: 1) CPU core speed has impact on both throughput and latency/jitter; 2) DPDK can allow performance prediction; 3) vhost-user is appropriate for real environment; and 4) OVS-DPDK provides a good combination ofperformance and functionality.