Slotted Flow Stats Acquisition: a Resource-efficient Flow Statistics Polling in OpenFlow Networks

Abstract

The Software Defined Networking (SDN) architecture becomes gradually more common in recent networking solutions considered in both research and commercial contexts. While SDN promises to simplify network management and monitoring processes, there are still numerous issues to address. Many of them involve the centralized network controller, depicted as a single point of failure able to make the whole network down in case of misconfiguration or unavailability. The SDN controller is typically involved in monitoring tasks that require uninterrupted connectivity with network nodes. Maintenance of up-to-date network state database generates frequent data transfers over the control channel and significant resource consumption. The both inevitably affect performance of other tasks. Considering such circumstances, it is desirable to find a tradeoff between measurements accuracy and the related overhead in terms of network and computational resources.In this paper, the Slotted Flow Stats Acquisition (SFSA) mechanism is proposed. Its aim is to spread in time the control channel load caused by network monitoring tasks, preventing the controller from being affected by resource consumption bursts. The mechanism may be adjusted to user-specific needs regarding desired accuracy of the periodically gathered measurement data. It is shown that the proposed solution effectively eliminates traffic spikes caused by requests for bulk flow statistics. Controller responsiveness in OpenFlow-based networks is thus improved.