The aim of the AQUILA project is to create, implement and evaluate a scalable enhanced end-to-end Quality of Service architecture for IP networks. During the ongoing project, the AQUILA project partners will observe the Internet2 initiative as well as IETF and other re-lated activities not only to be as open as possible to new developments, but also to actively contribute.

This deliverable collects an overview and identifies quality of service sensitive products as well as implementations which could be the base for the project at an early stage. First rec-ommendations for the project are given.

Both provider and user requirements will be summarised in order to be in the position to establish a customer friendly environment that is manageable and scalable as well. To meet the actual user requirements, questionnaires were designed for both end users and business customers. The subjectively felt insufficiencies of the current Internet were addressed as well as the preferred services and the access methods used.

Derived from the collection of information which are presented in this WP1.1 deliverable the following items are initial recommendations for the other WPs of the project:

• The starting point of the AQUILA architecture will be a light DiffServ solution. Light DiffServ means the reduction to very limited service classes which are characterised by a set of technical parameters and identified by their codepoints.
• Key services were identified and their traffic requirement characteristics evaluated to give guidance to WP 1.2 in defining the service classes. Agreement was obtained to support 2 to 4 service classes, where 2 is the system inherent minimum number of classes and 4 service classes were found to be sufficient to satisfy the key applications' needs. Each service class should have equal handling and an equal parameter set. A preliminary system of 4 service classes was proposed, consisting of the "Premium Voice" (for delay and jitter sensitive applications, e.g. Voice over IP), the "PremiumMul-tiMedia" (delay sensitive applications), the "PremiumMissionCritical" (loss sensitive ap-plications) and the "Standard" (traditional best effort) service classes.
• An important requirement for a provider is the ability to offer VPNs to business custom-ers, MPLS may be used for this purpose. The importance of MPLS is recognised, but MPLS will not be an essential part of AQUILA during the initial efforts. First trials MPLS exploitation focuses only on VPN awareness.
• RSVP is accepted only as a possible solution for access signalling, not for end to end res-ervations through the core network.
• In order to be able to offer, guarantee and account for the offered services, the provider needs measurements of the achieved performance in each of the classes. For this purpose, distributed QoS measurement systems per service class will be developed and per-formed during two trials.
• For the trials solely DiffServ and QoS aware commercial router platforms available at the partners will be used. Policy-based QoS management is only in the early stages of de-velopment and the approach differs from vendor to vendor, but the policy-based idea will still be observed during the project as progress is made quite quickly. For the Resource Control Layer the use of SUN workstations is planned. Hardware analysers/generators as well as measurement software tools appropriate for the trials are commercially available. Specialised software for IP tuning i.e. traffic shapers can be applied to network elements.
• Prior to the first trial, performance tests of the available edge routers will be performed. The router software releases will be evaluated regarding their QoS features. As most of the available equipment is CISCO-made, the various IOS releases have to be checked.
• QoS IP software components, software development platforms, simulation packages, and tools for measurement were investigated. On the level of standard PC operating systems, support for QoS is implemented in Linux and Windows2000. The project will focus on a platform independent solution. JAVA (1.2.x), with availability of a CORBA imple-mentation, was chosen as the most promising platform for software development as it al-lows operating system independent applications.
• Regarding middleware, although there is a huge choice, CORBA was found to be the most promising. It seems to be the right choice for the implementation of the resource control agents (RCA, EDA) and the End User Application Toolkit interface in AQUILA.
• UML will be used for/during the analysis and design activities of the project (use case view, logical view...). The preferred CASE (Computer Aided Software Engineering) tool is Rational Rose.
• QoS routing will not be part of the first trial.
• The initial efforts (1st trial) will not deal with inter-domain issues.
• Development of specific hardware is not part of the project.