JANET(UK) statement about QoS on JANET

The investigation of QoS on JANET is an important area of the JANET Development Programme. This statement outlines the current policy and position of QoS on JANET.

Summary of the key points Traffic on the JANET backbone will not be subject to QoS treatment as the backbone infrastructure is provisioned with sufficient headroom in terms of both bandwidth and router capacity to avoid congestion and therefore traffic delay or loss, effectively meaning that all traffic is treated as if it were IP Premium. JANET(UK) also has the capability of adding additional bandwidth and router capacity in a responsive manner to maintain sufficient headroom on the backbone as network traffic levels grow. To accommodate any special requirements such as the transfer of large point-to-point traffic flows across JANET, dedicated capacity can be provisioned using the JANET Lightpath service. The JANET backbone will be QoS transparent, allowing any QoS marked IP traffic to pass through the network unchanged. Where it is not feasible or economic to provide enough bandwidth and router headroom, Regional Network Operators and JANET-connected organisations may deploy QoS on their networks, subject to requirements, and the JANET backbone will transport the markings used for end-to-end QoS without alteration but will take no action on marked packets. Regional Network Operators who deploy QoS on their network will be required to follow the recommended JANET QoS standards developed by the project, which will be published by JANET(UK) during Q3/2007. Specific advice and recommended standards on deploying and transporting traffic with QoS markings across JANET will also be provided for JANET-connected organisations and other networks that connect to and use JANET (such as the schools networks). JANET(UK) has published the QoS Technical Guide to help JANET-connected organisations and Regional Network Operators recognise network performance related problems in order to determine how best to resolve them. Going forward, QoS will continue to be an important area of the JANET development programme, ensuring that JANET(UK) keeps abreast of ongoing work in the standards bodies and the industry in general and that it also remains active in understanding how developments in QoS techniques can support services on JANET in the future.

Background - the JANET QoS Project

The QoS Development Project began in 2002, to assess the benefits of QoS on JANET in a multi-domain and multi-vendor environment for a range of production applications. Phase 1 of the project followed the recommendations made in July 2001 by the JANET QoS Think Tank, which conducted an in-depth review of the QoS technologies and standards, and drew together the expertise of specialists within the JANET community. The major recommendation made by the Think Tank was to investigate an approach based on the DiffServ standards developed by the IETF.

During Phase 1, the JANET backbone and participating regional and campus networks were configured for QoS in order to conduct large scale testing of three levels of service: the IP Premium service, the standard Best Effort service (default transport service of IP networks) and Less than Best Effort service for non-urgent bulk traffic. The trials that were carried out showed some obvious benefits in using different levels of service for a number of applications, such as the IP Premium service for popular delay-sensitive applications, like Videoconferencing and Voice over IP, over periods of network congestion on low bandwidth connections. The tests also revealed that provisioning and monitoring QoS would be major challenges.

Phase 2 of the project was established in November 2005 primarily to investigate the feasibility of defining and deploying QoS in production for the JANET community. The focus of Phase 2 was to develop a generic Quality of Service model which would be stable, consistent and maintainable across the JANET core and regional networks.

Several possible QoS models were discussed and investigated during the first year of Phase 2. In parallel, the project participants carried out three JANET-wide trials to further investigate the practical aspects of QoS benefits specifically for a range of popular applications in use within the JANET community. By December 2006, sufficient work had been carried out by the project participants to determine the complexities involved in maintaining an operationally stable QoS service across the multi network domain of JANET, and as a result the project and JANET(UK) concluded that pervasive QoS treatment will not be established on JANET in the near future. The regional and campus network operators will have an option to increase the level of bandwidth and router capacity and/or deploy QoS on their networks, where required. The JANET backbone will remain transparent to QoS (DiffServ) marking: DSCP values used for any end-to-end QoS service will remain unaltered by the backbone, and no action will be taken in the backbone to apply admission control, policing, shaping, scheduling or any other per-hop behaviour to packets.

The following are the main reasons for not implementing QoS treatment pervasively on JANET.

The level of complexity in provisioning and maintaining stable and consistent QoS to provide some guaranteed limits on delay and packet loss which can be bound within a service level agreement is high. In a multi-domain (multiple AS) context, domain-level state is needed to support end-to-end provisioning, in particular to prevent inter-domain (valid or malicious) traffic contention and congestion destroying QoS performance. Currently, no standardised production-quality components to support this have been identified.

The JANET core network currently has sufficient bandwidth and router capacity headroom to provide packet delay and loss characteristics at current traffic levels which are acceptable for real-time (as well as data) traffic on JANET. Traffic analysis suggests that the capacity available is likely to remain sufficient throughout the lifetime of the network. However, this will be kept under close review and if anticipated capacity levels are exceeded, JANET(UK) will revisit the requirements for QoS on JANET.

During the remaining term of the QoS project (until the end of November 2007) the project participants will produce recommendations and technical guidance focusing on the edges of JANET, in particular access links between regional networks and JANET-connected organisations, where QoS or enhanced provision is more likely to be required because of capacity limitations. A standards framework on how Regional Network Operators and JANET-connected organisations could deploy QoS on their networks, should they need to do so, will also be produced.

Although current demand for QoS services within the JANET community is low, as highlighted by the Network Performance Survey that JANET(UK) conducted in March 2006. [Results] JANET(UK) will continue to maintain QoS as an important area of the JANET Development Programme, ensuring that JANET(UK) keeps abreast of ongoing work in the standards bodies and the industry in general, and that it also remains active in understanding how developments in QoS techniques can support services on JANET in the future. JANET(UK) plan to track the ongoing work of the GEANT2 project which has developed the QoS service framework for the GEANT2 network. The NHS network has the potential to deploy a number of QoS classes on their network. Any mapping of DSCP values between the NHS network and JANET will be subject to further study.

JANET(UK) invites discussion of all QoS-related requirements, problems and suggestions with the community. Please feel free to contact the JANET(UK) project manager, Victor Olifer (v.olifer@ja.net, tel.: 01235 822243) direct or join and email the QoS discussion mailing list, qos-forum@jiscmail.ac.uk.