Minimizing Transient Congestion during Network Update in Data Centers

• Published in: Proceedings - International Conference on Network Protocols pp. 1 - 10
• Main Authors: Zheng, Jiaqi, Xu, Hong, Chen, Guihai, Dai, Haipeng
• Format: Conference Proceeding Journal Article
• Language: English
• Published: IEEE 01.11.2015
• Subjects: Algorithm design and analysis; Algorithms; Approximation algorithms; Congestion; Data centers; Networks; Optimization; Planes; Protocol (computers); Routing; Switches; Switching theory; Transient analysis; Utilization; Wide area networks
• ISSN: 1092-1648
• DOI: 10.1109/ICNP.2015.33

To maximize data center network utilization, the SDN control plane needs to frequently update the data plane as the network conditions change. Since each switch updates its flow table independently and asynchronously, the state transition -- if done directly from the initial to the final stage -- may result in serious flash congestion and packet loss. Prior work strives to find a congestion-free update plan with multiple stages, each with the property that there will be no congestion independent of the update order. Yet congestion-free update requires part of the link capacity to be left vacant and decreases utilization of the expensive network infrastructure. Further, it involves solving a series of LP, which is slow and does not scale well. In this paper, we study the more general problem of minimizing transient congestion during network update, given the number of intermediate stages. This exposes the tradeoff between update speed and transient congestion, and allows an operator to navigate a broader design space for performing network update. We formulate the minimum congestion update problem (MCUP) as an optimization program and prove its hardness. We propose an approximation algorithm and a greedy improvement algorithm to find the update sequence in an efficient and scalable manner. Extensive experiments with Mininet show that our solution reduces update time by 50% and saves control overhead by 30% compared to state of the art.