Title: Towards a transmission control mechanism for data centres based on flow deadline and size over software defined networking
Authors: Yu Tianfang; Qiu Xuesong
Addresses: School of Computer Science, State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Haidian, Beijing, China ' School of Computer Science, State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Haidian, Beijing, China
Abstract: Data centres (DCN) support lots of online services, which possess strict deadline constraints. Some research shows that TCP protocol cannot provide deadline-aware transmission services for intra-DCN traffic, because TCP protocol cannot learn about upper application information and always fairly allocate bottleneck bandwidth. As a result, those throughput-sensitive flows have bigger sending windows, yet deadline-aware flows have to queue for forwarding along with longer delays. To address the issue above, we present DSATC, a transmission control mechanism based on flow deadline and size. DSATC uses double ECN thresholds to mark packets, takes the remaining completion time and size of a flow to calculate the flow urgency and redefines the adjustment method of the congestion window (CWND) to balance congestion control and flow deadline completion rate. The experimental results show that DSATC not only efficiently raises flow deadline completion rate and reduces flow completion time but also lifts network throughput.
Keywords: data centre; SDN; deadline-aware; congestion control; TCP protocol.
DOI: 10.1504/IJIPT.2023.133030
International Journal of Internet Protocol Technology, 2023 Vol.16 No.3, pp.148 - 160
Received: 27 Jun 2022
Accepted: 25 Dec 2022
Published online: 25 Aug 2023 *