Special Issue on Multimedia Storage Systems Richard Muntz Computer Science Department UCLA Los Angeles, CA 90095-1596 rnun tz@cs, ucla. edu Multimedia information content is growing fast and becoming a vital part of nearly all information communication and presentation. It is quickly becoming imperative to have a compelling multimedia presentation to attract potential customers to your web page or to keep the attention of your audience at a presentation. Multimedia brings with it a new set of parameters and constraints that provide the challenge of revisiting the issues of storage, search, delivery, presentation, etc. that have been addressed over the years in other contexts. This is a large area of research and in scope it is well beyond the capacity of one issue of PER. This issue is focused particularly on storage system issues. The articles in this issue were invited from some of the most active researchers from industry and academia in this area and address the major issues in multimedia storage systems. The first article by Ozden, Rastogi and Silberschatz titled "Architecture Issues in Multimedia Storage Systems" provides an overview of the design issues for next generation storage systems and sets the stage for the remaining papers which focus on more specific topics. In "Buffer Sharing in Video-onDemand Servers" Shi and Ghandeharizadeh explore ways of exploiting the sequential nature of video to share buffer resources between multiple delivery streams of the same video and thus reduce disk access requirements. Golubchik, in the paper titled "On Issues and Tradeoffs in Design of Fault Tolerant VOD Servers" addresses the question of how to provide the most cost effective fault tolerance. The issues are different in continuous media systems as compared to non real-time data servers. In essence, without real-time requirements, a failure may just cause degraded performance, e.g., a bit lower mean response time. For a continuous media system, the options have to be more precisely preplanned, e.g., leaving slack bandwidth to absorb reduced capacity under failure, or having a multilevel encoding scheme with different levels of QoS which have different bandwidth requirements. The forth article, by Muntz, Santos and Berson, describes the design and measurement of a parallel disk storage system that provides a stochastic guarantee for the maximum latency of I/O requests while maintaining a relatively high utilization. The proposed design can handle all types of multimedia as well as non real-time workload components. The final paper, "Adaptive TTL Schemes for Load balancing of Distributed Web Servers" by Colajanni and Yu, studies the problem of balancing the load on web servers by directing traffic from clients to the less loaded server that can satisfy a request. The issues addressed include heterogeneous server capacity as well as the time varying nature of the client request traffic. I hope that you enjoy this issue of PER and find it both informative and interesting. Finally, my thanks to the invited authors for their excellent contributions!

/lp/association-for-computing-machinery/special-issue-on-multimedia-storage-systems-0iuKyi0qKP