SDN-Based Multi-Class QoS Guarantee in Inter-Data Center Communications

Abstract

In this paper, we describe a software defined networking (SDN) approach to traffic engineering in inter-data center communication; then present <inline-formula><tex-math notation="LaTeX">$\mathrm{MCTEQ}$</tex-math></inline-formula> , a constrained utility-optimization formulation of the joint-bandwidth allocation problem for multiple classes of traffic in inter-data center communication. <inline-formula><tex-math notation="LaTeX">$\mathrm{MCTEQ}$</tex-math></inline-formula> handles priorities between traffic classes in a soft manner and explicitly considers the delay requirement of <inline-formula><tex-math notation="LaTeX">$\mathrm{Interactive}$</tex-math></inline-formula> flows. <inline-formula><tex-math notation="LaTeX">$\mathrm{MCTEQ}$</tex-math></inline-formula> being NP-hard, we construct a new approximation to be able to lean on the mature and efficient linear programming solvers to obtain fast and accurate approximations. We demonstrate via numerical experiments with two realistic inter-data center network topologies that <inline-formula><tex-math notation="LaTeX">$\mathrm{MCTEQ}$</tex-math></inline-formula> achieves considerably more network utilization than the best known solutions from the literature, while running at least twice faster. We also show via ns-2 packet level simulation that with <inline-formula><tex-math notation="LaTeX">$\mathrm{MCTEQ}$</tex-math></inline-formula> traffic engineering, end-to-end packet delay requirements of <inline-formula><tex-math notation="LaTeX">$\mathrm{Interactive}$</tex-math></inline-formula> flows are indeed guaranteed for in-profile traffic, which is not the case for alternative approaches.