C. COURCOUBETIS, V. SIRIS, AND R. WEBER. Investigation of cell scale and burst scale effects on the cell loss probability using large deviations. In: Proceedings of the 13th UK workshop on performance engineering of computer and telecommunication systems, Ilkley, United Kingdom, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....the loss curve, see our results on small buffers [17] and the large buffer asymptotics found in [2] and [16] The present paper focuses on the shape of the loss curve in the context of a mixed packet burst model. To our knowdledge, the only study that pursues this approach is Courcoubetis et al. [3]; they however derive only a few partial results. In more detail, the contributions are the following: i) In the combined packet burst model, locally the number of sources in the on state is constant and the model reduces to a number of (strictly) periodic streams that are multiplexed. The ....

....buffer length bcrit. At the same time, the optimizing T from the definition of I1 jumps from 0 to some positive integer. We do not have an explicit expression for bcrit it should be evaluated numerically. Notice that this value is essentially different from the value found by Courcoubetis et al. [3]. They claim that bcrit is the b such that = The rationale behind this heuristic is the following. As is clear from Theorem 3.1, the left hand side of the equation represents the decay rate for small values of b. The right hand side is explained by the fact that, in order for the phase ....

[Article contains additional citation context not shown here]

C. COURCOUBETIS, V. SIRIS, AND R. WEBER. Investigation of cell scale and burst scale effects on the cell loss probability using large deviations. In: Proceedings of the 13th UK workshop on performance engineering of computer and telecommunication systems, Ilkley, United Kingdom, 1997.

....used in that paper to implement WF scheduling can be adapted to implement algorithms with the complexity of the FSC. However due to its complexity, implementations of this algorithm are still difficult. Examining the timescale between packet level scheduling and burst level scheduling 52 [Courcoubetis97] among others [Roberts92, Chapter 8] noted the importance of packet rather than burst level scheduling as well as an examination of the traffic conditions that cause one effect or the other to dominate the multiplexing buffer. The importance of timescales is clear, and other authors (e.g. ....

C. Courcoubetis, V. A. Siris, and R. Weber. Investigation of cell scale and burst scale effects on the cell loss probability using large deviations. In Proceedings of the 15th UK Workshop on Performance Engineering of Computer and Communication Systems (UKPEW'97), Ilkley, UK, July 1997. 254

....scale, and accurately computes the cell loss probability. This contrasts with other work which has addressed either the cell scale or the burst scale alone. Using standard ideas from large deviations theory, we are able to investigate the qualitative nature of cell loss for very small buffers [CSW97] We show that there is a critical buffer size above which cell scale effects are no longer active, in which case it is valid to adopt the fluid model. However, when the amount of buffer per source is very small, both burst and cell scale effects are present (see Figure 3.3) The burst scale ....

C. Courcoubetis, V. A. Siris, and R. Weber. Investigation of cell scale and burst scale effects on the cell loss probability using large deviations. In Proc. of the 15th UK Workshop on Performance Engineering of Computer and Communication Systems (UKPEW'97), Ilkley, UK, July 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC