Effects of Early Update Cost in Cloud Object Storage Scheduling

Abstract

© 2021, Springer Nature Switzerland AG. Object storage, a type of cloud storage as a service, allows organizations to have highly available storage with zero infrastructure investments and thus fewer operation costs. Usually, a fixed-selection of storage tier on the cloud service is not appropriate since the workload on the objects is not static, and each storage tier cost can be drastically different from another. To efficiently select the appropriate object storage tiers, tedious work is usually required for planning, which could be modeled as the job shop scheduling problem with a pre-calculated threshold. To the best of our knowledge, no solution considers an early update cost, a penalty cost when the objects are moved out to another tier or deleted before an agreed stored-time is reached. In this paper, we present the effect of the early update cost of the existing cloud object scheduling algorithms, particularly when objects in a cold tier are read heavily, or spike read. From the experiments, it is found that the performance of the algorithms can be enormously worse when the EUC is applied.