An approach of collecting performance anomaly dataset for NFV Infrastructure

EasyChair Preprint 494

Abstract

Today, Network Function Virtualization(NFV) technology is widely used in industry and academia. At the same time, it presents a lot of challenges to reliability, such as anomaly detection, anomaly location, anomaly prediction and so on. All of these studies need a very large number of anomaly data information. This paper designs a method for collecting anomaly data from IaaS, and constructs a anomaly database for NFV applications. Three types of anomaly data sets are created for anomaly study, includes workload with performance data, fault-load with performance data and violation of Service Level Agreement(SLA) with performance. In order to better simulate the abnormality in the production environment, we use Kubernetes to build a distributed environment, and to accelerate the occurrence of abnormality, a fault injection system is utilized. Our aim is to provide more valuable anomaly data for reliability research in the NFV environment.

Keyphrases: Anomaly database, Clearwater, Fault Injection., IaaS, Kubernetes, NFV, performance monitoring

@booklet{EasyChair:494,
  author    = {Qingfeng Du and Yu He and Tiandi Xie},
  title     = {An approach of collecting performance anomaly dataset for NFV Infrastructure},
  doi       = {10.29007/8b62},
  howpublished = {EasyChair Preprint 494},
  year      = {EasyChair, 2018}}