Characterizing and Orchestrating NFV-Ready Servers for Efficient Edge Data Processing
Characterizing and Orchestrating NFV-Ready Servers for Efficient Edge Data Processing
复制标题
DOI：
10.1145/3326285.3329057
复制
发表时间：
2019-06
期刊：
影响因子：
--
通讯作者：
Lu Zhang;Chao Li;Pengyu Wang;Yunxin Liu;Yang Hu;Quan Chen;M. Guo
中科院分区：
文献类型：
--
作者：
Lu Zhang;Chao Li;Pengyu Wang;Yunxin Liu;Yang Hu;Quan Chen;M. Guo
快速增长的物联网(IoT)和人工智能(AI)应用要求高性能的边缘数据分析。关注小型架构(例如加速器)或大型基础设施(例如云数据中心)的先前工作无法完全满足这一要求。位于边缘和云之间，已经有许多服务器级别的设计用于增强边缘数据处理。然而，它们通常需要专门的硬件资源，并且缺乏可扩展性和灵活性。除了重新发明轮子外，我们还探索在即将到来的5G时代利用未充分利用的网络基础设施来增强边缘数据分析。具体地说，我们专注于在启用网络功能虚拟化(NFV)的商用服务器上高效部署边缘数据处理应用。通过这种方式，我们可以受益于NFV的服务灵活性，同时大大降低了部署在边缘网络中的许多服务器的成本。我们通过大量的实验研究了基于DPDK的NFV平台中数据包处理的特点，发现了使用DPDK轮询模式时存在的资源未充分利用问题。然后，我们提出了一个名为EdgeMiner的框架，该框架可以获取内核潜在的空闲周期，用于数据处理。同时，当虚拟网络功能(VNFS)和边缘数据处理(EDP)应用在同一服务器上共同运行时，它还可以保证它们的服务质量(Qos)。
The fast-growing Internet of Things (IoT) and Artificial intelligence (AI) applications mandate high-performance edge data analytics. This requirement cannot be fully fulfilled by prior works that focus on either small architectures (e.g., accelerators) or large infrastructure (e.g., cloud data centers). Sitting in between the edge and cloud, there have been many server-level designs for augmenting edge data processing. However, they often require specialized hardware resources and lack scalability as well as agility. Other than reinventing the wheel, we explore tapping into underutilized network infrastructure in the incoming 5G era for augmenting edge data analytics. Specifically, we focus on efficiently deploying edge data processing applications on Network Function Virtualization (NFV) enabled commodity servers. In such a way, we can benefit from the service flexibility of NFV while greatly reducing the cost of many servers deployed in the edge network. We perform extensive experiments to investigate the characteristics of packet processing in a DPDK-based NFV platform and discover the resource under-utilization issue when using the DPDK polling-mode. Then, we propose a framework named EdgeMiner, which can harvest the potentially idle cycles of the cores for data processing purpose. Meanwhile, it can also guarantee the Quality of Service (QoS) of both the Virtualized Network Functions (VNFs) and Edge Data Processing (EDP) applications when they are co-running on the same server.