Effeclouds: A cost-effective cloud-of-clouds framework for two-tier storage

Abstract Cloud storage is attracting more and more users with its virtually infinite capacity. However, users have to confront challenges of service outages, price hikes and vendor lock-in. Data replication is a widely used approach to overcome these challenges, but it may bring reduplicated growth in costs. Tiered-storage, which springs up in recent years, offers an opportunity for cost optimization. The hot tier charges users a higher price of storage but a lower price of operations, and vice versa for the cold tier. For optimizing costs, users can transfer their data between hot and cold tiers according to the prior knowledge of data’s future access patterns which, however, is generally very hard to obtain in practice. Therefore, it is cost-risky to transfer data rashly because data would incur more costs if they are stored in inappropriate tiers. To achieve cost-effectiveness and high data availability, in this paper we propose a cost-effective framework, Effeclouds, which leverages erasure coding to stripe data across multiple clouds redundantly, in addition to an online algorithm with proved competitiveness to further optimize costs by selecting appropriate tiers for data. Effeclouds offers users a simple way to deploy and exploit, and simultaneously gives consideration to both cost-effectiveness and data-availability. It also mitigates the effect of vendor lock-in and price hikes on costs with multiple clouds as the back-end. Eventually, we validate the effectiveness of Effeclouds by extensive simulations driven by real-world traces.