Congestion Control Mechanisms for Data Center Networks

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence

Title: "Congestion Control Mechanisms for Data Center Networks"

By

Mr. Wei BAI


Abstract

In recent years, many data centers have been built around the world to 
provide various services to global users. Many applications within the 
data center have very demanding latency requirements. Even a small delay 
due to the network congestion can directly affect application performance 
and degrade user experience. Therefore, how to handle the congestion in 
data center networks is very important.

The thesis focuses on congestion control mechanisms for data center 
networks. Specifically, we make the following three contributions.

First, we present PIAS, a flow scheduling mechanism to minimize the flow 
completion time. PIAS does not assume the prior knowledge of the flow size 
information and can be easily implemented using commodity switch hardware 
and legacy network stacks.

Second, we show that existing Explicit Congestion Notification (ECN) 
marking schemes suffer from various performance impairments in 
multi-service multi-queue data centers. Then we propose MQ-ECN, a new ECN 
marking scheme for round-robin schedulers that are widely used in today's 
data centers. Further, driven by recent progress in programmable 
schedulers, we design TCN, which can enable ECN for arbitrary packet 
schedulers in data centers.

Third, we show that existing transport solutions suffer from either 
excessive packet losses or low throughput in high-speed extremely 
shallow-buffered data center networks. Then we propose BCC, a simple yet 
effective solution with only one more ECN configuration at the switch. BCC 
maintains low packet loss rate persistently while keeping high throughput 
until the buffer becomes insufficient.


Date:			Friday, 19 May 2017

Time:			2:00pm - 4:00pm

Venue:			Room 2302
 			Lifts 17/18

Chairman:		Prof. Hai Yang (CIVL)

Committee Members:	Prof. Kai Chen (Supervisor)
 			Prof. Wei Wang
 			Prof. Qian Zhang
 			Prof. Jiang Xu (ECE)
 			Prof. Fengyuan Ren (Comp. Sci. & Tech., Tsinghua
 					    Univ.)


**** ALL are Welcome ****