AuthorsC. Griwodz and P. Halvorsen
EditorsA. Tveito, A. M. Bruaset and O. Lysne
TitleRELAY - on the Performance and Resource Utilization of Time-Dependent Large-Scale Distributed Systems
AfilliationCommunication Systems, Communication Systems
StatusPublished
Publication TypeBook Chapter
Year of Publication2009
Book TitleSimula Research Laboratory - by thinking constantly about it
Chapter17
Pagination199-236
PublisherSpringer
Place PublishedHeidelberg
ISBN Number978-3642011559
Abstract

Powerful end-systems and faster access networks make large-scale distributed multimedia applications such as streaming and gaming increasingly popular. This popularity does, of course, also mean that the increase in available computing power and network bandwidth does not solve all existing problems, but rather that other parts of distributed systems will become bottlenecks. To relieve these, RELAY addresses issues of resource utilization in such distributed systems by investigating network and operating system support that provides quantifiably better support than the state-of-the-art. One of our main goals is to integrate and combine mechanisms to get more scalable, less resource demanding, high performance systems for time-dependent large-scale distributed multimedia systems. RELAY designs, implements and evaluates algorithms, mechanisms and tools to improve resource utilization, increase throughput, reduce/hide latency and support soft Quality-of-Service. The development takes user experience into account, since adaptation specific to this can reduce the necessary resource consumption considerably. We consider architectural, application-level, kernel and protocol support to achieve our goals, and we investigate combinations of performance-enhancing mechanisms such that they do not counteract each other. RELAY's approach is application-driven, where the best way of achieving performance improvements determines the research methodology. As an example, we report several of our findings in investigating support for distributed interactive applications, and massive multiplayer games in particular.