This paper deals with reducing costs for cloud computing users. Cloud customers use “Traffic Redundancy Elimination” (TRE) to reduce bandwidth costs. Redundant data chunks are detected and removed – cloud providers will not implement middleboxes for this as they have no incentive. The paper gives a TRE solution which does not require a server to maintain client status. The system is known as PACK “Predictive ACKnowledgements” which is receiver driven.
The investigation of network traffic using statistical analysis to gain insight into performance and behaviour.
This paper looks at the issue of reducing variability in performance in data centre networks. Variable network performance can lead to unreliable application performance in networked applications – this can be a particular problem for cloud apps. Virtual networks are proposed as a solution to isolate the “tenant” performance from the physical network infrastructure. The system presented is known as Oktopus. The system provides a tradeoff between guarantees to tenants, costs to tenants and profits to providers by mapping a virtual network to the physical network.
This paper looks at RTT and how they affect TCP flows. Because of limited data they look at how to derive RTTs from analysis of tcpdump data (unidirectional).
Methods of getting estimates of RTT are:
SYNSYN-ACKACK – the time stamp between SYN and ACK in triple handshake.
Flight method – look at packets with near identical inter-packet times. Calculate time between start of flights (attempt to ignore rate-limited flows).
This paper considers the problem of balancing traffic across network egresses. It achieves a workable solution using a scalable packet market scheme which couples end-hosts controlling their own connection with an overall controller which can select routes appropriately for each flow.
The flows are balanced to seek paths which minimise loss.
This paper looks at the phenomenon of long-range dependence. It shows that certain long-range dependent models give answers which contain infinities and also that this behaviour will not be detected by a naive modelling approach. The work is extended in the JCCS paper
Talk to Next Generation Networks meeting -- this talk is partly based on the ICC paper (fundamental relationships in TCP flows) and shows how TCP throughput arises from network parameters and engineering decisions.