| Traffic shaping mainly aims to resolve the network traffic congest as result of traffic burst. With the rapid development of Internet, network services vary quickly and the number of uses is also increasing. How to distinguish the network traffic more explicitly and how to provide unified administration and hierarchical scheduling towards the traffic stream of different users and network services so as to assure the rationally use of the network resources turn into a noticeable problem.This paper puts forward MTRA(Multilayer Token Re-Allocation) traffic shaping algorithm which is suitable to run on DiffServ model aiming at solving the problem that traditional traffic shaper cannot dynamically adjust the shaping parameter and the self-adjusted shaper is to difficult to implement. The key idea of MTRA algorithm is using a multilayer token reallocation model to reallocate the surplus tokens generated by the token buckets and to some extend supply different network services dynamic bandwidth. Thus the different network stream can share the bandwidth more reasonable and flexible. At last, the feasibility of the model has been analyzed by mathematical means.The key part of the MTRA algorithm is the token reallocation rule by which the surplus token is allocate. As the network structure gets more and more complicated, network services classification also gets more and more explicit and the user population surges rapidly, the paper conceives the token reallocation algorithm. According to the network condition, the algorithm uses different reallocation rules to dispatch the surplus token and thus achieve the goal of dynamic allocating the network bandwidth. The validity of the reallocation rules has been proved by experiment.At last, experiment was made on reconfigurable router and the capability of the algorithm was tested. The differences between MTRA traffic shaping and traditional traffic shaping was profoundly analyses thus the validity and practicability of MTRA was proved. |
PDF Full Text Request