Resource Optimization for Service Chain Monitoring in Software-Defined Networks

Emerging services enabled by software-defined networking (SDN) and network function virtualization (NFV) are introducing new challenges to network monitoring. One such network service is "service chaining", where the complexity of the data path forwarding is coupled with the heterogeneity of service functions. This makes it extremely difficult to monitor and troubleshoot networks with a reasonable resource overhead. We propose a framework that can be applied for service chain monitoring. A monitoring hierarchy is defined, which converts a monitoring intent to monitoring zones and probes for on-demand monitoring. To minimize the cost of probe deployment, we design a heuristic algorithm that conducts dynamic pruning and search. We also identify the possibility for monitoring-rule consolidation when multiple probes are co-located, and introduce the use of Bloom Filters to optimize monitoring resource usage. Our performance evaluation shows reduced probe-deployment cost by our Tree Search algorithm in various settings, and significant memory savings by rule consolidation.