Modelling and Optimizing Bandwidth Provision for Interacting Cloud Services

Chao Chen, Ligang He, Bo Gao, Cheng Chang, Kenli Li, Keqin Li

Abstract: Non-deterministic communication patterns among interacting Cloud services impose a challenge in determining appropriate bandwidth provision to satisfy the communication demands. This paper aims to address this challenge and develops a Communication Input-Output (CIO) model to capture data communication produced by Cloud services. The proposed model borrows the ideas from the Leontief’s Input-Output Model in economy. Based on the model, this paper develops a method to determine the bandwidth provision for individual VMs that host a service. We further develop a Communication-oriented Simulated Annealing (CSA) algorithm, which takes an initial VM-to-PM mapping as input and finds the mapping with the minimal bandwidth provision and without increasing the PM usage in the initial mapping. Experiments have been conducted to evaluate the effectiveness and efficiency of the CIO model and the CSA algorithm.