Efficient traffic matrix estimation for data center networks

We address the problem of estimating the real-time nature of traffic flows in data center networks, using the light-weight SNMP data. Unlike the problem of estimating the traffic matrix (TM) across origin-destination (OD) pairs in ISP networks, the traffic flows across servers or ToR (Top of Rack) switch pairs in data center networks are notoriously more irregular and volatile. Although numerous methods have been proposed in past several years to solve the TM estimation problem in ISP networks, none of them could be applied to data center networks directly. In this paper, we make the first step to decompose the data center topology to several clusters by leveraging the characteristics of prevailing data center architecture, which makes TM inference problems in data center networks easy to handle. We also state a lemma to obtain the coarse-grained traffic characteristics of these clusters unbiasedly. Two efficient TM inference algorithms are proposed based on the decomposed topology and the coarse-grained traffic information, which improves the state-of-the-art tomography methods without requiring any additional instrumentation. Finally, comparing with a recent representative TM inference algorithm through intensive simulations, the results show that, i) the data center TM inference problem could be well handled after the decomposition step, ii) our two algorithms outperforms the former one in both speed and accuracy.