Adaptive Measurement-based QoS Management in DiffServ Networks for Variable Bit Rate Applications
1st International Workshop on Interdomain Performance and Simulation (IPS ’03)
DiffServ, Linear Prediction,, QoS Manager, Token Bucket Filter
In per-flow based resource reservation techniques, bandwidth allocation for Variable Bit Rate (VBR) applications through real time flow measurements is limited. This limitation is due mainly to scalability problems. In aggregated traffic flow networks such as Differentiated Services (DiffServ), measurement-based bandwidth allocation leads to optimisation of network resources as it links resource allocation to actual usage. With Constant Bit Rate (CBR) applications, simple summation of pre-specified bandwidths is a sufficient solution for resource optimisation in DiffServ networks. However, the pre-specified bandwidth technique is an inefficient approach to DiffServ bandwidth allocation for VBR applications. These applications are generally unaware in advance of their specific bandwidth requirements. This paper describes an adaptive Quality of Service (QoS) management technique that provides an efficient resource allocation scheme for VBR applications in aggregate traffic flows. This method is, in effect, a measurement-based and linearly predicted bandwidth allocation scheme.