Routing and Scheduling Issues in Support of Advance Reservations

点击翻译按钮获取中文摘要

The general topic of the proposal is advanced reservations in itegrated packet networks. The ability to secure in advance the resources needed by a distributed application is becoming increasingly important. This is true, not only for high-end scientific or medical applications which require simultaneous access to several facilities and pieces of sophisticated equipment, but also for more mundane applications such as video-conferencing, whose reasonably high resources requirements make it desirable to ensure that the necessary network resources are available at the same time as the conference participants are. Support for advance reservations involves multiple aspects such as extensions to protocols, and in particular reservation (signalling) protocols, and enhanced resource management procedures, e.g., call admission. While there has been a number of previous works addressing both of these aspects, the corresponding two sets of issues have typically been considered separately. In addition, the advance reservation models that have been used, have essentially focused on asingle service model, i.e., the availability of a given amount of resources, from a given starting time, and for a given duration (with possibly a random component in both the starting time and the duration). The goals of this proposal are, therefore , to investigate jointly protocol extensions and enhancements to resource management, as well as consider more general service models for advance reservations. In particular, the researchers will consider the problem of routing of advance reservation requests in the context of an extension to the OSPF protocol, where link state updates have been extended to carry resource availability infromation over different time intervals. This allows for a distributed implementation which improves scalability by avoiding that all advance reservation requests be forwarded to a central server. However, it introduces new problems in order to contain the corresponding increase in protocol traffic while ensuring adequate performance. In that context, the researchers will study issues related to the trade-off between information accuracy and routing performance, and also investigate the use of mechanisms such as preemption priorities to offer applications a range of service options, e.g., request large amount of bandwidth at a low priority together with a small amount at a high priority. In addition, the researchers will also consider, again jointly with the associated routing problem, various extensions to the advanced reservation model, such as mininmizing completion time, maximizing available bandwidth, etc., that are of potential interest to adaptive applications that can modify their behavior according to the available service. The researchers believe that the joint development of the necessary infrastructure, e.g., extensions to the OSPF protocol, and the investigation of desirable extensions to the traditional advance reservation model, will allow them to make significant progress in both our basic understanding of the inherent complexity in supporting advance reservation, and our ability to deploy such capabilities.