|
To satisfy the ultra-low latency requirement of cloud services, a lot of congestion control mechanisms have been proposed to reduce the Flow Completion Time (FCT) in data center networks.Theoretically, SRTF could achieve minimal FCT. However, existing congestion control mechanisms either do not achieve SRTF or are difficult to be deployed.This paper analyzes the challenges of achieving global-optimal Shortest Remaining Time First (SRTF) scheduling in a congestion control mechanism. Then a credit-based congestion control mechanism, Helm, is proposed. Helm solves the challenges by carefully combining the finite priority queues at switches and infinite rate setting at receivers and thus achieves near global-optimal SRTF. With theoretically analysis, Helm can achieve near global-optimal SRTF. Besides, extensive simulations are conducted and the results show that Helm reduces the mean and tail FCT by up to 62\% and 75\%, respectively, compared with Homa. |
|
Keywords:Computer Architecture; Data Centers; Congestion Control; SRTF; Low Latency |
|