Open Collections

On profile optimization in DOCSIS 3.1 networks exploiting traffic information

University of British Columbia

The latest release of the Data over Cable Service Interface Specification (DOCSIS), namely DOCSIS 3.1, has introduced the use of adaptive bit loading profiles. The optimal design of these profiles is critical for the overall performance of a Hybrid Fiber Coaxial (HFC) cable system. In this thesis, we introduce a low complexity profile design algorithm that maximizes the average throughput of a HFC system by grouping its cable modems (CMs) on the basis of both their downstream data arrival rates and channel signal-to-noise ratios (SNRs). Optimizing the profiles on the basis of both CM traffic and CM channel SNRs is challenging as these two criteria are somewhat contradictory. To account for the CMs' channel SNRs, CMs with similar SNR distributions should be grouped into the same profiles. However, to take into account the CMs' traffic, CMs with high data arrival rates should be grouped into profiles with higher bit loading capabilities, and CMs with lower data arrival rates should be grouped into profiles with lower bit loading capabilities.%, irrespective of their channel SNRs. In this thesis, we take into account these two contradictory criteria in the profile optimization problem, and we propose a sub-optimal solution to solve this problem. Performance analysis results of our algorithm show that accounting for both CM channel SNR and CM traffic can improve the average throughput of a HFC system. Furthermore, our approach accounts for the CMs' traffic without requiring future knowledge of their actually realized data arrival rates, which is difficult to obtain in practice. We also consider an asymptotic scenario, where the per-profile arrival rates (which is the sum of the arrival rates of all the CMs in a profile) is always guaranteed to be less than the respective per-profile bit loading capacities. For this case, our algorithm optimizes the system profiles with the objective of minimizing the system's total average data transmission time. We show that our proposed approach yields significantly high performance, particularly for scenarios with a high degree of heterogeneity in the CMs' traffic.

Text

2018-11-05

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/67747

Electrical Engineering

University of British Columbia

UBCO

http://creativecommons.org/licenses/by-nc-nd/4.0/