Improved Flow Awareness By Spatio-Temporal Collaborative Sampling In Software Defined Network

With the rapid increase in the number of Internet users and devices connected to the Internet,these dramatic changes in network size places a huge demand for flexible and scalable network management.General traffic analysis based on deep packet inspection techniques at the switches cannot grasp the detailed knowledge of network applications going into internal switches,and the statistics-based reports of switches are also lack of flow-level recognition of the traffic.Therefore,network-wise accurate flow-awareness by packet sampling is highly desirable for fine-grained quality of service guarantee,internal network management,traffic engineering and so on.In order to improve sampling accuracy as much as possible under limited sampling resources and sampling nodes,we propose a Spatial-Temporal Collaborative Sampling(STCS)framework in the flow-aware softwaredefined networks(SDNs).In particular,taking into account the spatial-temporal factors and limits of network resources,the goal of the formulated STCS problem is to maximize the network-wise sampling accuracy of flows including mice and elephant flows by characterizing both of the comprehensive influences of switches and the effects on sampling accuracy imposed by the collaborative strategy.The comprehensive influences combin the current influences brought by traffic dynamics of the switch and the potential influences quantified by the spatial location and historical activity.STCS constructs a cooperative method of staggering sampling in the time dimension by the relationship of switches in each spatial position through the flow path,which maximizes the overall influence of the system.Meanwhile,the staggered manner minimizes redundant packets.Finally,we propose a suboptimal approach to address the complex STCS problem in two steps: 1)Top-K switch selection based on iterative comprehensive influence,and2)sampling time slot allocation based on the local value maximization.In summary,this paper builds a testbed environment based on real network topology and implements the STCS solution framework.We evaluate the proposed approach by a realistic largescale topology,and the results show that the sampling accuracy can be effectively improved by the method,especially for mice flows,and the redundant ratio of sampled packets is reduced to less than 10%.