An Efficient Algorithm for Runtime Minimum Cost Data Storage and Regeneration for Business Process Management in Multiple Clouds

Abstract

The proliferation of cloud computing provides flexible ways for users to utilize cloud resources to cope with data complex applications, such as Business Process Management (BPM) System. In the BPM system, users may have various usage manner of the system, such as upload, generate, process, transfer, store, share or access variety kinds of data, and these data may be complex and very large in size. Due to the pas-as-you-go pricing model of cloud computing, improper usage of cloud resources will incur high cost for users. Hence, for a typical BPM system usage, data could be regenerated, transferred and stored with multiple clouds, a data storage, transfer and regeneration strategy is needed to reduce the cost on resource usage. The current state-of-art algorithm can find a strategy that achieves minimum data storage, transfer and computation cost, however, this approach has very high computation complexity and is neither efficient nor practical to be applied at runtime. In this paper, by thoroughly investigating the trade-off problem of resources utilization, we propose a Provenance Candidates Elimination algorithm, which can efficiently find the minimum cost strategy for data storage, transfer and regeneration. Through comprehensive experimental evaluation, we demonstrate that our approach can calculate the minimum cost strategy in milliseconds, which outperforms the exiting algorithm by 2 to 4 magnitudes.