Details

Autor(en) / Beteiligte

• Moscholios, Ioannis D.
• Logothetis, Michael D.
• Nikolaropoulos, Periklis I.

Titel

Engset multi-rate state-dependent loss models

Ist Teil von

• Performance evaluation, 2005-02, Vol.59 (2), p.247-277

Ort / Verlag

Elsevier B.V

Erscheinungsjahr

2005

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• Call blocking probabilities (CBP) are a key index of quality of service assessment in multi-rate networks. We focus on the CBP calculation in single link loss models where calls of each service-class come from finite sources and compete for the available link bandwidth under the complete sharing policy. We review the Engset multi-rate loss model (EnMLM) and the single-retry loss model for finite sources (f-SRM) in which blocked calls of a service-class may immediately retry once, in order to be connected in the system, with reduced bandwidth and increased service time requirements. We extend the f-SRM to the multi-retry loss model for finite sources (f-MRM) to cover multiple retrials. Furthermore, we propose the single- and multi-threshold finite source models (f-STM, f-MTM) in which calls, prior to being blocked, may adjust their traffic requirements according to a common set of thresholds, which indicate the occupied link bandwidth. Moreover, we generalize the above-mentioned models with the connection-dependent threshold model for finite sources (f-CDTM), in which each service-class may have its own set of thresholds. The proposed models do not have a product form solution and therefore we resort to approximate but recursive formulas for the calculation of the link occupancy distribution, which is essential for the CBP calculation. The link occupancy distribution calculation is more complex than that in the corresponding infinite source models, since it requires enumeration and processing of the state space. To this end, we propose a new method for the state space determination, applicable to the proposed models. Furthermore, we present the f-CDTM for service-classes of a mixture of finite and infinite sources. Finally, we evaluate the models’ accuracy by comparing the analytical with simulation CBP results.