Understanding the characteristics of Internet traffic and designing an efficient RaptorQ-based data transport protocol for modern data centres

Alasmar, Mohammed (2019) Understanding the characteristics of Internet traffic and designing an efficient RaptorQ-based data transport protocol for modern data centres. Doctoral thesis (PhD), University of Sussex.

[img] PDF - Published Version
Download (19MB)


This thesis is the amalgamation of research on efficient data transport protocols for data centres and a comprehensive and systematic study of Internet traffic, which came as a result of the need to understand traffic patterns and workloads in modern computer networks.

The first part of the thesis is on the development of efficient data transport pro- tocols for data centres. We study modern data transport protocols for data centres through large scale simulations using the OMNeT++ simulator. We developed and experimented with an OMNeT++ model of NDP. This has led to the identification of limitations of the state of the art and the formulation of research questions with respect to data transport protocols for modern data centres. The developed model includes an implementation of a Fat-tree topology and per-packet ECMP load bal- ancing. We discuss how we integrated the model with the INET Framework and validated it by running various experiments that test different model parameters and components. This work revealed limitations of NDP with respect to efficient one-to-many and many-to-one communication in data centres, which led to the de- velopment of SCDP, a novel and general-purpose data transport protocol for data centres that, in contrast to all other protocols proposed to date, natively supports one-to-many and many-to-one data communication, which is extremely common in modern data centres. SCDP does so without compromising on efficiency for short and long unicast flows. SCDP achieves this by integrating RaptorQ codes with receiver-driven data transport, in-network packet trimming and Multi-Level Feed- back Queuing (MLFQ); (1) RaptorQ codes enable efficient one-to-many and many- to-one data transport; (2) on top of RaptorQ codes, receiver- driven flow control, in combination with in-network packet trimming, enable efficient usage of network re- sources as well as multi-path transport and packet spraying for all transport modes. Incast and Outcast are eliminated; (3) the systematic nature of RaptorQ codes, in combination with MLFQ, enable fast, decoding-free completion of short flows. We extensively evaluated SCDP in a wide range of simulated scenarios with realistic data centre workloads. For one-to-many and many-to-one transport sessions, SCDP performs significantly better than NDP. For short and long unicast flows, SCDP performs equally well or better compared to NDP.

In the second part of the thesis, we extensively study Internet traffic. Getting good statistical models of traffic on network links is a well-known, often-studied problem. A lot of attention has been given to correlation patterns and flow duration. The distribution of the amount of traffic per unit time is an equally important but less studied problem. We study a large number of traffic traces from many different networks including academic, commercial and residential networks using state-of-the-art statistical techniques. We show that the log-normal distribution is a better fit than the Gaussian distribution. We also investigate a second, heavy- tailed distribution and show that its performance is better than Gaussian but worse than log-normal. We examine anomalous traces which are a poor fit for all tested distributions and show that this is often due to traffic outages or links that hit maximum capacity. Stationarity tests showed that the traffic is stationary at some range of aggregation times. We demonstrate the utility of the log-normal distribution in two contexts: predicting the proportion of time traffic will exceed a given level (for link capacity estimation) and predicting 95th percentile pricing. We also show the log-normal distribution is a better predictor than Gaussian orWeibull distributions.

Item Type: Thesis (Doctoral)
Schools and Departments: School of Engineering and Informatics > Informatics
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5101 Telecommunication > TK5105.5 Computer networks
Depositing User: Library Cataloguing
Date Deposited: 15 Jan 2020 15:20
Last Modified: 15 Jan 2020 15:20
URI: http://sro.sussex.ac.uk/id/eprint/89371

View download statistics for this item

📧 Request an update