Cloudifying Mobile Network Management: Performance Tests of Event Distribution and Rule Processing

Abstract

With the ever increasing number of devices, nodes and the events they create, scalability and performance become important aspects for Operation Support Systems (OSS). One solution is to distribute the work load, i.e. ‘cloudify’ the formerly centralized monitoring and decision functions. This requires remodeling Complex Event Processing (monitoring) and Policies (decision making) towards a distributed yet coordinated system. This paper describes an extended architecture, implementation and performance tests for a policy-based event processing system. The main advantage of our approach is that we use policies for event pattern matching (an advanced form of Complex Event Processing) and for the selection of corrective actions (called Distributed Governance). Policies are (a) distributed (over multiple components) and (b) coordinated (using centralized authoring). The resulting system can deal with large numbers of incoming events, as is required in a telecommunication environment. Peak load will be well above 1 million events per second, combining different data sources of a mobile network. This paper presents the motivation for such a system, along with a comprehensive presentation of its design, implementation and evaluation.