Monitoring, aggregation and filtering for efficient management of virtual networks

Abstract

Abstract—Virtual Networks are characterised as highly dynamic network environments, where topologies and nodes adapt rapidly to changes in user and service demands, user location and context changes, or resource constraints. This paper presents a rigorous assessment of an Information Management Overlay (IMO) suitable for monitoring virtual networks. The IMO system allows complex monitoring tasks to be performed in a scalable manner in order that the network manager can monitor the desired network properties in a way which is accurate, scalable, and low-bandwidth. The monitoring architecture is decentralised, not resource intensive (in terms of memory or CPU), and adaptable to a wide range of virtual topologies. The IMO architecture uses Information Aggregation Points and Information Collection Points to scalably aggregate, filter, and collect data. The IMO is tested for scalability in both simulation (with over 35,000 thousand nodes) and on a custommade virtual network testbed (with over 700 virtual routers), where we show that it performs well at a variety of monitoring tasks. This paper shows that the system is scalable and that intelligent configuration of the monitoring system greatly improves its efficiency. Furthermore, filtering can help reduce the monitoring load, however, the improvement brought in by filtering is greatly dependent on the nature of the monitoring task. I.