As part of a broader organisational restructure, data networking research at Swinburne University of Technology has moved from the Centre for Advanced Internet Architecture (CAIA) to the Internet For Things (I4T) Research Lab.

Although CAIA no longer exists, this website reflects CAIA's activities and outputs between March 2002 and February 2017, and is being maintained as a service to the broader data networking research community.


Broadband Internet Traffic Simulation & Synthesis (BITSS)

Overview

An emerging consumer broadband context is a home LAN with one or more PCs. Potentially a network-attached game console and VoIP adaptors(s) all sharing an ADSL or DOCSIS access link. The consumer's broadband link (today varying from 128Kbit/sec to 20Mbit/sec) is a key congestion point between the consumer's 100MBit/sec or 1Gbit/sec home LAN and the ISP's networks.

Sevice quality as perceived by the ISP customer comes from the combination of simplying the end-user's interaction with their broadband access devies and engineering trade-offs between available IP bandwidth, available mechanisms for differential queuing and scheduling of competing IP traffic classes at congestion points. In an ideal world, the end-user's VoIP, video, online games and related services 'just work'. In the real world, interactive application data flows, such as VoIP and online games, are disrupted in various ways by protocols that induce packet loss or latency in order to measure the network's available capacity (typically TCP-based applications - web browsing, email, peer to peer content delivery and son on).

A key challenge for ISPs is to characterise and predict the impact of traditional TCP-based services on the experienced quality of emerging VoIP and IP video services in the broadband consumer space. A related challenge is to predict and understand how the growth of highly interactive online environments (such as real-time multiplayer games) will impact on, and be impacted by, VoIP, IP video and traditional TCP-based applications.

VoIP service quality is currently a `hot topic` for ISPs looking to manage customer experience of IP-based triple-play offerings. Furthermore, interactive multiplayer online games represent a rapidly growing market where degradation or fluctuation of network service quality is quickly noticed and criticized by players. Forward-looking ISPs have already begun hosting game servers on their own backbones to improve customer experience. Understanding the impact of thousands or tens of thousands of ADSL-attached or DOCSIS-attached consumers playing interactive games will be crucial over the coming decade.

Modelling of multiplayer game traffic will primarily focus on the first person shooter (FPS) genre. FPS games have been shown to be both QoS-sensitive (network layer latency, packet loss and jitter has a noticeable degradation on customer satisfaction) [1] and dissimilar to VoIP (utilising noticeably asymmetric streams of variable sized packets between clients and servers) [2]. Although FPS game servers often support less than ~64 players, there are tens of thousands of such game servers active around the internet at any one time, hosted by enthusiasts, ISPs and dedicated game hosting companies. Our modelling will focus both on synthesising diurnal and weekly variations in bulk traffic load, and the packet-by-packet patterns (in time and packet size variations) during active game-play.

Other multiplayer game genres, such as online car racing games, and massively multiplayer role playing games (MMORPGs) or strategy games will also be explored in controlled lab trials. Role playing and strategy games do not traditionally exhibit much sensitivity to normal latency fluctuations [3], but are beginning to incorporate FPS-like elements within game play. However, unlike FPS games the MMORPG traffic model tends towards thousands of clients connecting to a common (possibly clustered) set of servers. Racing games are likely to exhibit latency-intolerance quite similar to that of FPS games, but have not been studied in great detail yet.

VoIP traffic patterns will be synthesised from literature on the topic and, if possible, from anonymised traffic traces provided voluntarily by VoIP-enabled ISPs.

Last Updated: Thursday 23-Jul-2009 16:28:38 AEST | Maintained by: Carl Javier (cjavier@swin.edu.au) | Authorised by: Grenville Armitage ( garmitage@swin.edu.au)