Information Technology Reference
In-Depth Information
way to the client's display, and employ resource allocation together with admission control in
these system components to guarantee performance.
To support resource allocation and admission control, we first need to establish performance
models of the underlying system components that can relate the workload to the desired per-
formance metric (e.g., delay, buffer requirement, etc.). Using the performance models the
admission controller can then compute the performance metric using the new workload (as-
suming a newmedia session is admitted, for example) to see if the required performance is still
met. If so, then the resources are allocated and the new media stream is admitted. Otherwise
the new media stream is rejected for service, or put into a queue to wait for sufficient resources
to become available.
In practice, however, modeling the system components is often far from simple, especially if
the system component has dynamic interactions with other system components. Thus, another
approach to admission control called observational admission control has been proposed in
the literature [3]. In observational admission control the system performance model is not
known (or not accurately known). The admission controller simply measures the resource
utilization as well as various performance metrics of the system, and uses these measurement
results to estimate whether admitting a new media stream will result in overload. Clearly, this
observational approach will not be able to provide a deterministic guarantee. Nevertheless, for
complex system components that cannot be modeled accurately, this observational approach
can still improve an otherwise entirely best effort service.
1.9 Summary
In this chapter we have reviewed the basic concepts of continuous media streaming, from
the types of media data, media delivery, to the essential system components. Moreover, we
presented the many new challenges in the design and implementation of continuous media
streaming systems, as well as the common dimensions for engineering trade-offs. Finally, we
discussed three types of performance guarantees and their relation to resource allocation and
admission control. In the rest of the chapters in Part I we will present more detailed discussions
on several key topics in continuous media streaming, namely, media data coding, compression,
and adaptation (Chapter 2), storage and retrieval (Chapters 3, 4), fault tolerance (Chapter 5),
and media data streaming (Chapters 6, 7, 8).
References
[1] M. Claypool and J. Tanner, The Effects of Jitter on the Perceptual Quality of Video, Proceedings of the Seventh
ACM International Conference on Multimedia , vol. 2, Florida, USA, Nov. 1999, pp. 115-118.
[2] B. Goode, Voice over Internet Protocol (VoIP), Proceedings of the IEEE , vol. 90, no. 9, Sept. 2002, pp. 1495-1517.
[3] D.D. Vleeschauer, J. Janssen, andG.H. Petit, Delay andDistortionBounds for PacketizedVoice Calls of Traditional
PSTNQuality, Proceedings of the 1st IP-Telephony Workshop (IPTel 2000) , GMDReport 95, pp. 105-110. Berlin,
Germany, 12-13 April 2000.
[4] D.A. Patterson, G.A. Gibson, and R.H. Katz, A Case for Redundant Array of Inexpensive Disks (RAID), Pro-
ceedings of the ACM Conference on Management of Data , Chicago, IL, USA, June 1988, pp. 109-116.
[5] H.M. Vin, A. Goyal, and P. Goyal, An Observation-based Admission Control Algorithm for Multimedia Servers,
Proceedings of the International Conference on Multimedia Computing and Systems , 15-19 May 1994, pp. 234-
243.
Search WWH ::




Custom Search