Towards Efficient Transports for Datacenter Networking with High Environmental Variations

PhD Thesis Proposal Defence


Title: "Towards Efficient Transports for Datacenter Networking with High 
Environmental Variations"

by

Mr. Junxue ZHANG


Abstract:

The datacenter networking is long believed to be stable and has little 
variations, therefore, the transports design for datacenter networking is 
based on such assumption. However, environmental variations actually exist 
in real-world datacenter networking. For instance, the RTT is assumed to 
be very stable in datacenters, however, due to the network components, eg, 
middlebox, hypervisor, etc, the RTT can have up to 2.68X variations. 
Furthermore, there are also other environmental variations in datacenters, 
posing challenges towards transports design for datacenter networks.

Regarding the RTT variations problem, it is difficult for datacenter 
operators to derive the proper ECN marking threshold to simultaneously 
deliver high throughput, low latency and good burst tolerance 
communications. Furthermore, we find that neural network driven transports 
can learn and adapt to the varying environment, which shows its potential 
to be successful in datacenter networking with high environmental 
variations. However, current neural network based transports suffer either 
performance loss or large overhead.

This thesis describes my research efforts in designing efficient 
transports for datacenter networking with high environmental variations. 
First, to solve the problem of degraded performance with high RTT 
variations, we propose ECN#, ECN# extends current ECN marking mechanism to 
consider both instantaneous and persistent congestion. Second, to make 
neural network based transports available for datacenter networking, we 
propose LiteFlow, LiteFlow breaks the tradeoff of either performance loss 
or increased overhead for those neural network based transports by 
automatically integrating a neural network in kernel datapath.


Date:			Wednesday, 11 August 2021

Time:                  	4:00pm - 6:00pm

Zoom Meeting: 
https://hkust.zoom.com.cn/j/98529261121?pwd=eGllSlBBNUNFY0lwUUhQZ1cvWlFSZz09

Committee Members:	Dr. Kai Chen (Supervisor)
 			Dr. Brahim Bensaou (Chairperson)
 			Prof. Bo Li
 			Prof. Qiong Luo


**** ALL are Welcome ****