Study Of The Implementation Of DiffServ PHBs

DiffServ, one of the solutions for Internet QoS, was put forward by IETF only several years ago. It possesses good scalability and robustness, and could be the framework of the future Internet (IPng). In this paper, our focus is on the key part of DiffServ, i.e., the implementation of Per Hop Behavior (PHB).Three kinds of PHBs have been specified by RFC documents. They are Expedited Forwarding PHB, Assured Forwarding PHB Group and Best Effort PHB. The implementation of PHB mainly involves buffer management and scheduling mechanism. The buffer management includes buffer allocation among different queues and buffer management in one physical queue. By use of computer simulation and mathematic analysis, we propose a new solution for PHB implementation. Our main contributions contain the following parts: 1) As for the buffer allocation among different queues, we put forward the DiffServ Buffer Allocation Scheme, based on a memory-sharing policy-sharing with maximum queue and minimum allocation. 2) The RIOC with staggered-parameters is selected for queue management. 3) PRI/FWRR+, a new two-level queuing mechanism, is designed for DiffServ scheduling. FWRR+ is based on WRR, but the two are quite different from each other. WRR can only be used in serving fixed size packets, while FWRR+ can also deal with variable size packets. In addition, FWRR+ is featured by good fairness and simplicity. Simulation tests indicate that our implementation of PHBs can satisfy the DiffServ requirements on network delay and link bandwidth. Finally, some further considerations are listed.