Data center traffic engineering using Markov approximation
Kouji Hirata, Miki Yamamoto · 2017
This paper proposes a data center traffic engineering scheme with Markov approximation. Markov approximation is a distributed optimization framework. It optimizes networks by independent behaviors of users that construct a time-reversible continuous-time Markov chain based on minimum necessary information in the networks. The proposed scheme aims at minimizing the maximum link utilization in data center networks by means of Markov approximation. The proposed scheme provides two strategies with different search space. Through simulation experiments, we show that the proposed scheme efficiently reduces the maximum link utilization under not only static situations but also dynamic situations where traffic demands dynamically change in data center networks.