Optimal Bandwidth Allocation for Data Transmission in Virtual Data Centers

Lisha Chen, Zejian Kang, Weining Dai, Zhemin Zhang, Cheng Wang, Jonathan Li

Abstract: In this paper, we propose to use the transmission time of data blocks as the metric to measure the network utilization, and study the bandwidth allocation for multi-path data transmission in virtual data centers. We start with the case that data transmission requires no minimum bandwidth, and formulate the bandwidth allocation problem as two convex optimization problems. In the first problem, the maximum transmission time of all data blocks is minimized under the given data block sizes and link capacities, which is named as the maximum transmission time minimization (MTTM) problem. The second convex optimization problem minimizes the transmission time summation of all data blocks, and the problem is named as the transmission time summation minimization (TTSM) problem. These two problems are then generalized to the case that data transmission has minimum bandwidth requirement, and the corresponding problems are named as general MTTM (G-MTTM) and general TTSM (G-TTSM) problems. By exploring the properties of the G-MTTM problem, we design a fast algorithm for it, named as TTM-based algorithm. Extensive simulation results demonstrate that the G-MTTM and G-TTSM problems can be efficiently solved, and to use transmission time of data blocks as the metric for network utilization measurement and optimization is practical.