A Study Of Lightweight Traffic Measurement In Software Defined Networks

With the rapid development of the Internet,the network traffic has continued to increase,the types of applications have been diversified,the network structure has become more and more complex,and the personalization of user needs has continued to increase.However,the rapid development of the network makes the traditional network architecture more difficult to expand,more complex equipment management,and higher resource redundancy.In view of the defects of traditional network architecture,Software Defined Network(SDN)makes it possible to simplify the network architecture,improve resource utilization and fine-grained network management by decoupling the control function and forwarding function on the switch.Fine-grained network management is based on fine-grained network traffic measurement.However,due to the high cost and high power consumption of Ternary Content Addressable Memory(TCAM),the flow table resources of the switch are very limited,leading to coarse-grained traffic measurement with flow table based traffic statistics mechanism.In addition,in a high-speed network,SDN switches are limited by limited computing resources.On the basis of ensuring packet forwarding functions,it is also a major challenge to achieve fine-grained traffic measurement.In general,limited by the computing resources and memory resources on SDN switches,the main research goal is to achieve lightweight and fine-grained traffic measurement using the characteristics of SDN control plane and data plane.The main contents and contributions of our research are presented as follows:·Considering the limited computing resources and memory resources on switchess and traffic distribution,we present a least-recently-used(LRU)list based solution,called LRU-TX,for lightweight and accurate per-flow traffic measurement.Different from the classic LRU list with one exit at the tail,we maintain another exit at a certain position r(not the tail)with a given threshold τ in LRU-TX list.Many terminated flows will be removed from the exit at position r as soon as possible with their counters less than the threshold τ,which helps to decrease the memory consumption for those terminated flows.We also formally analyze the impact of different parameters(e.g.,the position r and the threshold τ)on the measurement performance.The extensive simulation results show that LRU-TX can increase the throughput by about 30%at least,and reduce the memory cost by about 50%,compared with the state-of-the-art solutions under the same measurement error conditions.Moreover,we tested the LRU-TX on the DPDK platform to process packets for a line-rate network,i.e.,achieving the throughput of about 14.88Mpps.·When targeting multiple types of applications,the corresponding types of sketches can be deployed on each switch to satisfy the requirements for the traffic measurement.If each arrival packet is measured by all sketches on each switch(i.e.,without sketch configuration),it may lead to redundant traffic measurement and massive computing resource consumption.In order to solve the problem with different measurement requirements,we propose a sketch configuration framework for multi-type applications.As a case study about balancing the number of measured flows,we present an optimal sketch configuration for proportional fairness with per-switch computing capacity constraint(SCP),so that each sketch can measure enough flows without unduly restricting the number of flows measured by other sketches.Due to the NP-hardness,a greedy-based algorithm with approximation ratio 1/3 is presented,and its time complexity is analyzed.The simulation results show that the proposed algorithm can measure more than 30%91%flows,compared with the existing solutions with resource constraint.·Since each switch works independently,a flow may be measured by multiple switches along its route path,leading to a waste of computing resource and memory resource on the switch.Therefore,many works focus on efficient centralized traffic measurement load distribution strategy in a single-controller scenario,called collaborative traffic measurement.However,with the growth of network topologies and network traffic,each controller only manages a subset of switches in the network with multi-controller,in order to improves the processing capacity and robustness of the control plane.The paper proposes a collaborative traffic measurement load distribution strategy in a multi-controller scenario.By updating the measurement probability in each packet header,it is guaranteed that a flow can be measured by only one switch along its route path,which effectively supports the cooperative flow measurement strategy in a multi-controller scenario.Moreover,compared with the state-of-the-art works,the maximum communication load of the control plane is reduced by about 20%,and the maximum traffic measurement load among switches is reduced by about 50%.In conclusion,focusing on the fine-grained traffic measurement in SDN,our dissertation conducts in-depth research on the above mentioned three problems,and the proposed solutions have achieved superior performances from both theoretical and experimental perspectives,providing valuable reference for the practical application of the fine-grained traffic management.