Server-Assisted Traffic Measurement for Programmable Data Center Networks

The daily management of today's data centers relies on traffic measurement, which offers essential information for various tasks. The emergence of the programmable networking paradigm paves the way for implementing fine-grained and accurate traffic measurement in data center networks. However, due to the limitations of hardware resources in programmable switches, it is still very challenging to achieve fine-grained and accurate traffic measurement solely relying on switches. To reduce the resource consumption and implementation complexity of traffic measurement schemes on switches, as well as to enhance measurement performance, we investigate server-assisted traffic measurement schemes for programmable data center networks in this paper. We propose two server-assisted traffic measurement schemes: Traffic Measurement with Server-Assisted Large Flow Identification (TM-SALFI) and Traffic Measurement with Server-Assisted Large Flow Measurement (TM-SALFM), which allows servers to aid in large flow identification and measurement. We implement the proposed schemes in an experimental network consisting of servers, OVS-DPDK, and P4 hardware switches. The experimental results demonstrate that the TM-SALFI and TM-SALFM can achieve significantly higher measurement accuracy with acceptable overheads compared to the state-of-the-art.