Auction Pricing for Bandwidth across Multiple Datacenters with Traffic Engineering

Bandwidth between servers becomes valuable and congested for network users because of the gradual development of datacenter networks recently. The bandwidth is selling at a fixed price per GB in traditional methods. Traffic engineering technique is commonly used for route scheduling. The main shortcoming of traditional pricing strategies is the low usage rate of costly connections and mendacious transactions of strategic users. Auction principle is able to overcome shortcomings in these aspects, but the unacceptable efficiency in large-scale networks limits the usability. And, existing pricing strategies do not take the importance of data and transaction itself into consideration in scheduling. In order to take advantage of the auction principle and avoid the shortcomings, MAPS, a pricing system for bandwidth across multiple datacenters based on auction principle, is designed. MAPS combines auction pricing with traffic engineering technique in order to achieve high performance. MAPS aims to achieve considerable efficiency and provides fairness guarantee for normal bidders and against malicious bidders. Algorithms in MAPS are well-designed based on the formulation of the problem. Our evaluation is based on simulation and the result proves that MAPS performs well in network utility and time efficiency. Important transactions are prioritized and the value of data transferred is improved up to 60 percent.