Reducing the price of resource provisioning using EC2 spot instances with prediction models

Abstract

The increasing demand of computing resources has boosted the use of cloud computing providers. This has raised a new dimension in which the connections between resource usage and costs have to be considered from an organizational perspective. As a part of its EC2 service, Amazon introduced spot instances (SI) as a cheap public infrastructure, but at the price of not ensuring reliability of the service. On the Amazon SI model, hired instances can be abruptely terminated by the service provider when necessary. The interface for managing SI is based on a bidding strategy that depends on nonpublic Amazon pricing strategies, which makes complicated for users to apply any scheduling or resource provisioning strategy based on such (cheaper) resources. Although it is believed that the use of the EC2 SIs infrastructure can reduce costs for final users, a deep review of literature concludes that their characteristics and possibilities have not yet been deeply explored. In this work we present a framework for the analysis of the EC2 SIs infrastructure that uses the price history of such resources in order to classify the SI availability zones and then generate price prediction models adapted to each class. The proposed models are validated through a formal experimentation process. As a result, these models are applied to generate resource provisioning plans that get the optimal price when using the SI infrastructure in a real scenario. Finally, the recent changes that Amazon has introduced in the SI model and how this work can adapt to these changes is discussed.

Evaluating cloud interactions with costs and SLAs

“…For this purpose, they used a heuristic method to configure the cloud broker to optimize the virtual machine pricing of the cloud broker. Fabra et al [26] proposed a framework to generate price prediction models for Amazon EC2 Spot Instances. Prediction models are applied to generate optimal resource provisioning plans.…”

A dynamic task scheduler tolerant to multiple hibernations in cloud environments

“…To the best of our knowledge, the hibernation mechanism of the spot VMs is only discussed in our previous works [27,28] and in [10]. However, in [10], although the authors consider a scenario where hibernation-prone spot VMs can be used, they recognize that deadline constraints add complexity to the problem of resource provisioning, which should be evaluated in detail. But, no practice solution is presented or analyzed.…”

Backup or Not: An Online Cost Optimal Algorithm for Data Analysis Jobs Using Spot Instances

“…They construct a model and reversely engineer how prices are set and find the price may be generated most of the time at random from within a tight price range via a dynamic hidden reserve price mechanism. Fabra et al [24] design a framework which can classify the spot instance availability zones and then generate price prediction models adapted to each class for generating resource provisioning plans that get the optimal price. The bidding strategy based on online algorithm [25] is also proved by Guo et al to be able to achieve price optimization.…”