Time-domain modeling of batching under user interaction and dynamic adaptive piggybacking schemes
Ke Wang, Prithwish Basu, Thomas D. C. Little · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2001
Provision of Video-on-Demand (VoD) services may require high network bandwidth and server capacity for sustained periods of time. Aggregation schemes can be used to increase the supported customer population under the constraints of these resources, but due to system behavioral complexity, these schemes have been difficult to model in the large scale. We present in this paper a time-domain analysis modeling technique that yields satisfactory performance estimation results. We show that our analysis is analogous to finding the time response of linear systems using the well known convolution theorem. We model two aggregation schemes using this technique: (1) batching by time-out and (2) adaptive piggybacking employing Snapshot-RSMA. Both schemes are server based aggregation schemes, and their importance increases as pervasive computing, with possibly many capacity-limited devices (such as PDAs) connected as potential recipients of streaming VoD, becomes closer to reality. Since such schemes do not assume any end client capacity requirements, they are the logical candidates for enabling efficient VoD service in a pervasive computing environment. We delineate the requirements of the VoD system under which this modeling technique can be employed and also propose a sampling methodology which gives good estimation results when modeling becomes too complex mathematically.