Building scalable, resilient and high-performance NFV systems

Abstract

Over the past years, Network Function Virtualization (NFV) becomes an important research topic. NFV technology advocates replacing hardware network functions (NFs) with virtualized NFs running on standard virtualization platform, which can significantly reduce the deployment cost and management complexity of NFs.

Despite of the advantages, practitioners of NFV technology are still facing several important problems. First, some network services are preferred to be deployed across geo-distributed datacenters so that users can enjoy better service quality. However, existing NFV management systems suffer poor scalability when deploying NFV service chains across geo-distributed datacenters, as scaling across multiple datacenters requires non-trivial interactions between various components of an NFV management system. Second, resilience functionality, including failure resilience and flow migration, is of pivotal importance in NFV system. However, existing NFV systems implementing resilience functionalities usually have unsatisfactory performance due to heavyweight check-pointing and centralized control. Finally, some NFs need to contact external services to achieve advanced functionalities. For the best performance, these NFs are usually implemented using callback-based asynchronous programming method, which is complex and error-prone. Moreover, the increased implementation complexity brought by callbacks further limits the development of better NF software.

Regarding the three problems, this thesis proposes three resulting systems, including ScalIMS, NFVactor and NetStar, to tackle these problems in a systematic way.

ScalIMS designs a dynamic NFV service chain scaling system that is capable of scaling NFV service chains across geo-distributed datacenters, using IMS system as a case study. ScalIMS combines proactive and reactive approaches for timely, cost-effective scaling of the service chains, and is evaluated on a geo-distributed public cloud. The evaluation result reveals that ScalIMS can effectively scale multiple NFV service chains across geo-distributed datacenters for improved traffic quality and reduced resource consumption.

NFVactor is a novel NFV system that aims to provide lightweight failure resilience and high-performance flow migration by leveraging actor model to improve the parallelism of resilience functionalities, while the efficiency of the actor model is guaranteed by a carefully designed runtime system. Moreover, NFVactor achieves transparent resilience: once a new NF is implemented for NFVactor, the NF automatically acquires resilience support. The evaluation result shows that NFVactor achieves 10Gbps packet processing, flow migration completion time that is 144 times faster than existing system, and packet processing delay stablized at around 20 microseconds during replication.

NetStar is a new NF programming framework that brings the future/promise abstraction to the NF dataplane. NetStar simplifies asynchronous NF programming via a carefully designed async-flow interface that exploits the future/promise paradigm by chaining multiple continuation functions for asynchronous operation handling. NF programs implemented using NetStar framework mimic simple synchronous programs, but are able to achieve full flow processing asynchrony with good performance. The evaluation result demonstrates that NetStar can effectively simplify asynchronous NF programming by substantially reducing the required lines of code, while still approaching line-rate packet processing speeds.