SwitchAgg: A Further Step Towards In-Network Computing

Many distributed applications adopt a partition/aggregation pattern to achieve high performance and scalability. The aggregation process, which usually takes a large portion of the overall execution time, incurs large amount of network traffic and bottlenecks the system performance. To reduce network traffic, some researchers take advantage of the emerging programmable RMT (Reconfigurable Match-Action Table) switch to commit In-Network Aggregation. However, we argue that the architecture of RMT switch is not well-suited for In-Network Aggregation since it is designed primarily for implementing traditional network functions. In this paper, we first show the weakness of RMT switch in handling aggregation tasks, and then we propose SwitchAgg, which is an In-Network Aggregation system that overcomes the shortcomings of RMT. We evaluate the performance improvement we have gained from SwitchAgg. Our results show that, SwitchAgg can aggregate data at line rate and give a high data reduction ratio, which helps us cut down network traffic and alleviate pressure on server CPU. In the system performance test, the job completion time can be reduced as much as 44%.