Using branch-and-price to maximize redundant network utilization in cloud datacenter
Shuo Wang, Qiqi Wang, Hongjie Zhang, Jing Li · 2017
In cloud datacenter, work-conserving bandwidth provision benefits the network sharing of multiple virtual datacenters (VDCs). Traditionally, cloud networks are shared in a best-effort manner making it hard to reason about how network resources are allocated. While cloud computing providers offer guaranteed allocations for resources such as CPU and memory, they do not offer any concrete bandwidth for network resources. The lack of concrete network resources prevents tenants from predicting lower bounds on the performance of their applications. Prior works concentrate on efficient and scalable bandwidth allocation algorithms for VDCs. However, per-VDC redundant bandwidth is ignored, which is crucial to work-conserving bandwidth offering. In this paper, we design an efficient link mapping algorithm to maximize per-VDC redundant bandwidth allocation. Due to the large-scale characteristic of cloud datacenter, it would generate huge search space for finding maximum bandwidth allocation, thereby incurring long solving time. Hence, we first divide the whole allocation spaces into independent pods, and model the link mapping to multi-commodity flow problem, which is linear programming problem with large number of decision variables. Afterward coupled with branch-and-price algorithm, we solve the link mapping efficiently. Simulations demonstrate that our VDC allocation achieves per-VDC redundant network allocation and low time cost.