Optimal-Coherent And Adaptive Software Defined Inference Of Network Traffics (Occasion)

In this paper, the OCcASION framework for Traffic Matrix (TM) estimation in Software Defined Networks (SDN) is proposed where, first in the learning phase, the Optimal Observation Matrix (OOM) of SNMP link-loads is estimated. Then in the measurement and inference phase, the OOM is used to coherently find the minimum-norm estimate of the unknown TM. This framework is applied under both non-adaptive and adaptive scenarios. The performance of OCcASION framework is evaluated using synthetic and practical traffic traces with real network topologies. It is shown that, comparing with regular minimum-norm estimation, this framework can significantly improve the accuracy of the TM estimation; for example, on Geant network the estimation error is approximately reduced by 83%.