End systems are controlled by end users, who are personally interested in experiencing a good performance for all of their applications. The majority uses one of few operating systems, whose developers are eager to provide end users with a good performance. Incremental changes that are limited to end systems’ operating systems do not require the involvement of developers of network equipment or network providers and can be deployed in a short timeframe. ￼￼￼￼
Through access networks, Internet service providers connect end users to the Internet. Access network equipment is located at the user side of the access connection and can include both wired and wireless networks. Within access networks it is possible to identify and distinguish individual network flows to and from local end systems, and to differentiate between them. Due to the current resource distribution in the Internet, bottlenecks are often located in an access network, and large buffers in access network equipment are typical. Major protocol changes or the introduction of new transport layer protocols requires updates of security mechanisms that exist in access networks. Deploying new mechanisms that require access network changes involves hardware vendors as well, but vendor-specific solutions in access networks can precede standardisation.
The core network consists of networks owned by a multitude of entities. Routers in these networks experience a higher degree of traffic multiplexing, making it difficult to differentiate between flows or even types of flows. Also, flows will usually cross several administrative entities. Any change that requires coordination between several of these entities takes major standardisation efforts. It is preferable for new mechanisms that they can be deployed incrementally and do not require coordination between administrative entities.
RITE explores means to reduce end-to-end latency, which requires that the existing mechanisms of all three blocks are understood. The development of new mechanisms, however, can be limited to some of the blocks. Work package 1 and 2 of RITE distinguish between end-system only modifications (WP1), which have a high probability of deployment during the lifetime of the project, and more fundamental changes that are expected to impact the design of network equipment and the way in which it is used in deployed networks (WP2). These changes require at least modification to the configuration of operational equipment and may imply upgrade or replacement of equipment, as such they must be expected to require much longer to become implemented in the Internet.
Work package 3 handles testbed deployment and prototype of the most promising solutions from WP1 and WP2 for the RITE uses cases of online gaming, financial trading applications, and interactive video conferencing. In RITE, developed mechanisms will be prototyped on the Linux operating systems and evaluated. Well-proven mechanisms will be submitted to the Linux kernel mailing lists as patches. Network changes will be prototyped and tested using Linux as a router and also through the involvement of ISP and Hardware partners (BT and ALU). In addition to operating systems contributions, RITE has a strong group of IETF experts that will take our results to this standardisation body. Dissemination, exploitation and management is coordinated through Work packages 4 and 5.