A Reliable and Cost-Efficient Auto-Scaling System for Web Applications Using Heterogeneous Spot Instances

摘要

Cloud providers sell their idle capacity on markets through an auction-likemechanism to increase their return on investment. The instances sold in thisway are called spot instances. In spite that spot instances are usually 90%cheaper than on-demand instances, they can be terminated by provider when theirbidding prices are lower than market prices. Thus, they are largely used toprovision fault-tolerant applications only. In this paper, we explore how toutilize spot instances to provision web applications, which are usuallyconsidered availability-critical. The idea is to take advantage of differencesin price among various types of spot instances to reach both high availabilityand significant cost saving. We first propose a fault-tolerant model for webapplications provisioned by spot instances. Based on that, we devise novelauto-scaling polices for hourly billed cloud markets. We implemented theproposed model and policies both on a simulation testbed for repeatablevalidation and Amazon EC2. The experiments on the simulation testbed and thereal platform against the benchmarks show that the proposed approach cangreatly reduce resource cost and still achieve satisfactory Quality of Service(QoS) in terms of response time and availability.