Modeling of Scheduling Algorithms of Downstream Channel Bonding in DOCSIS3.0 Based on GSPN

摘要

For QoS of service flows, it is essential to study on multi-channel load balancing and scheduling problems in Data Over Cable Service Interface Specifications (DOCSIS3.0). In this article, we studied downstream multi-channel load balancing and scheduling in DOCSIS3.0 by mean of Generalized Stochastic Petri Net (GSPN). We presented the model for two queuing mechanisms including the output queuing architecture and the input queuing architecture using GSPN, and described load balancing algorithms including Random Selecting (RS) and Shortest Expected Delay (SED), packets scheduling algorithms including Queue Length Threshold (QLT), Deficit Round Robin (DRR) and Bonded DRR (BDRR) based on GSPN model. In order to improve the delay performance of real-time service flow, we proposed Queue Threshold DRR (QTDRR) algorithm. Then we analyzed and evaluated SED-DRR.SED-QTDRR, RS-BDRR, RS-QTDRR on the basis of the theory and tool of GSPN. Our results show that QTDRR algorithm can improve the delay characteristic of real-time flows without affecting throughput.