LEVERAGE News 4

LEVERAGE
LEarn from Video Extensive Real Atm Gigabit Experiment

LEVERAGE News No 4, July 1998

Welcome | Co-ordinator | Pedagogical | Keynote conference address | Native ATM | Conference report | Acronym key

Technophile
Ramon Felder, research engineer at ASCOM Tech in Bern, Switzerland describes the challenge of finding new networking options for the forthcoming international trial
Adapting the LEVERAGE network for the third international user trial
Communication between local trial sites in different countries is central to the main aim of the LEVERAGE project which is to develop and demonstrate network-based co-operation in a multimedia multilingual environment. The JAMES project, which facilitated ATM connections for most of the transnational ACTS projects, provided a 10Mbps ATM link between the University of Cambridge (CULC) and the Institut National de Télécommunications (INT) near Paris during the preparation and running of the second LEVERAGE trials in September to November 1997. The termination of the JAMES project in March 1998 which could not have been foreseen at the start of the LEVERAGE project has therefore produced an unexpected challenge for our networking engineers.
As the LEVERAGE consortium is determined to complete the third phase of the project despite this contraint, alternative solutions for intersite connectivity during the third and final trial have had to be investigated and evaluated.
The third user trial will link all three sites, namely: CULC in the UK, INT in France and for the first time the Universidad Politécnica de Madrid (UPM). It will thus involve: three languages; three academic institutions and three local networks. The trial is scheduled to take place during November and December of 1998 and will prepare for the transfer of the ideas, methodologies and technologies developed during the LEVERAGE project into post-project applications.
Interconnection options
The IP API employed by the system is a prerequisite for the portability of applications onto a variety of networking infrastructures. A brief 'tour d'horizon' of available services, commercial and other, reveals four interconnection options:
Internet / TEN-34;
ISDN;
Private (leased) lines; and
Frame Relay (FR).
It is inevitable that the available bandwidth for intersite connection will be reduced for the third trial. In the case of an Internet solution, the lower bandwidth and unpredictable performance of the network would require us to lower the bandwidth available to the system significantly. With other options cost is a serious concern.
For the second trial bandwidth availability was a minor problem and imposed few restrictions. The video connections were configured for high quality consuming 800kbps, including audio. For this together with other applications and adding IP and ATM overhead a generous 2Mbps were allocated per session. With up to 5 groups of users conducting sessions in parallel this resulted in a total requirement of 10Mbps for the interconnection bandwidth.
Bandwidth reduction
Under the new constraints of availability and cost, the LEVERAGE consortium has been forced to consider ways of using far less bandwidth. However, bandwidth reduction will also be an important step towards commercially viable applications of LEVERAGE-like systems. To achieve the goal of minimal bandwidth usage we have two sets of variables to play with:
(i) the organisation of the trials: the number of groups participating; the number of parallel sessions and the number of sites involved in one session;
(ii) the bandwidth requirements of the individual communications streams generated by the LEVERAGE co-operation tools: mainly videoconferencing; but also audioconferencing and text-based tools.
If we accept that the pedagogical concept of peer tutoring in language learning works most effectively with a minimum of two languages involved, we can limit the trial sessions to pairs of sites. In other words, at any one time only two of the three sites will exchange communication streams between them (CULC and INT, CULC and UPM or INT and UPM). This simplifies networking as well as system aspects and offers a potential reduction of interconnection bandwidth (and hence, cost) depending on the technology selected.
In contrast to the fixed connection time-slots offered by the JAMES ATM link, all of the alternative connection options are either available on a permanent basis or on short term (dial-up) request. We therefore have far more flexibility for the scheduling of trial sessions. If only one group involving users from two sites is using the system at any one time bandwidth requirements are lower although this will also provide extra challenges for the pedagogical partners when planning and managing the sessions for the trials.
If we now look at the communication streams, videoconferencing requires the bulk bitrate. The results of the first trial which were conducted locally on the Cambridge site seemed to show that high quality audio was a greater priority for the users than high quality videoconferencing, however, the second trial has shown that in non-native/native speaker interaction the video is as important as the audio. Good quality is therefore imperative.
The H.261 ¹ realtime video encoding and decoding hardware employed in the LEVERAGE system allows reduction of the generated bitrate to a minimum of 64kbps. Tests were conducted locally at the INT site to define the minimum bitrate yielding an 'acceptable' image quality. Acceptance was subject to end-user criteria such as lip-movement and lip-synch. The result, due to improvements made over the last year, was a surprisingly low 200kbps. The audiostream is fixed at 64kbps and will be maintained to ensure good voice quality. Therefore, including the IP and ATM overhead and other tools, a session could be run with just 320kbps of available bandwidth!
Juggling the connections
With this in mind we can start evaluating interconnection options with a target bandwidth between 384kbps and 1Mbps. If possible we will maintain the option of having parallel sessions and/or enhanced video quality. We can now have a closer look at the options set out above.
Internet, TEN-34
The Internet is the Internet, TEN-34 is an IP network linking a number of European universities and research facilities. The main drawback of any IP network when used for realtime applications, is the lack of service quality guarantees. Although tests of the TEN-34 network at UPM have yielded a throughput of 400kbps for the UPM-INT link, which would be sufficient to sustain a LEVERAGE session, there is no way of telling how this value will fluctuate over time.
Since the trial will again form part of students' courses and good quality communications are crucial for the evolution of the system and successful co-operation of the participants, this uncertainty is unacceptable. This reduces the Internet and TEN-34 to a fallback solution.
ISDN
The European digital telephone network allows dial-up connections at 64kbps. Multiple connections can be aggregated to provide n*64kbps. Once the connections are set up the communication channel is transparent.
Private (leased) lines
These permanent circuits are provided with capacities of n*64kbps. The channel is transparent and can be used for data or voice communication. The choice between ISDN and leased lines is usually based on a purely economic analysis of the break-even between fixed and variable costs of both solutions.
Frame Relay
Originally deployed as pure data networks, standards and equipment are now available to provide voice/data integration over FR. This also makes FR a valid alternative for multimedia applications. The packet-oriented nature of the technology allows network resources to be shared between users, and hence a reduction of cost. Service quality issues such as throughput, delay and delay variation need further investigation.
As this issue of LEVERAGE News goes to press we are favouring a FR solution, which seems to offer the best cost/performance ratio. However, our investigations will continue and we are looking forward to the third trials when our plans and calculations will be put to the test.
¹ H.261 is the encoding/decoding standard employed by standard H.320 and H.323 videoconferencing applications.