Non-deterministic Optimal Pricing of VMs in Cloud Environments: An IGDT-based Method

Abstract

Today, cloud markets, especially Amazon, have attracted a lot of attention from users due to the provision of Spot Virtual Machines (SVMs). It has several advantages for both sides of the market. On the one hand, Amazon can generate revenue from its underutilized virtual machines. On the other hand, the customer can get the SVM as needed at a dynamic price through an auction method. Providing optimal bidding strategies in such a market is a crucial challenge. The bidding price is affected by uncertain parameters such as the price of SVMs, the number of available SVMs, the number of current customers, and their bidding values. In this paper, we use Information Gap Decision Theory (IGDT) to determine the best bidding strategy. Our proposed method includes both risk-averse and risk-neutral strategies. The evaluation results based on historical Amazon EC2 prices confirm the effectiveness of the proposed method in the presence of uncertain prices. It has high performance compared to the baseline methods in terms of robustness cost, uncertainty budget, and execution time.