Dynamic Provisioning for Differentiated Services on the Internet

Despite the growing abundance of bandwidth, a key problem of ISPs today is to package their bandwidth resources into high-value end-to-end services (SLAs). We propose a dynamic provisioning building block which can be a part of future edge-based services architecture supporting shortterm point-to-anywhere assured bandwidth contracts. Specifically, we develop a dynamic bandwidth allocation technique for a class-based services network. The algorithm uses an adaptive Exponentially Weighted Moving Average (EWMA) whose weight parameter adapts to the correlation horizon of the underlying long-range dependent traffic stream. The second order statistics collected at the estimated correlation horizon is used as the basis for bandwidth provisioning. The performance of the algorithm is illustrated using a simulated network and a policy-based framework. Keywords— Dynamic provisioning, Differentiated Services, Bandwidth Allocation