Abstract
Currently, Storage-as-a-Service (StaaS) clouds offer multiple data storage and access pricing options which usually consist of hot and cold tiers. The cold tier storage option offers a lower storage price while the hot tier storage option offers a lower access price. Cloud users need to choose an optimal tier to store their data objects economically based on the frequency of accesses to their data objects. Besides, StaaS cloud users can transfer data objects between these two tiers to save cost according to the varying frequency of accesses to their data objects. Therefore, in order to make optimal transferring decisions, future access curves are needed to be predicted. However, for cloud users, it is difficult to precisely predict future access frequencies for their data objects. In this paper, we propose an online algorithm to guide StaaS cloud users in making decisions on whether and when to transfer their data objects between cold and hot tiers for achieving cost optimizations, while users do not need to have any prior knowledge of future access frequencies. We prove theoretically that the proposed online algorithm can achieve guaranteed competitive ratios for data objects stored in a two-tier StaaS cloud. Finally, through extensive experiments, we validate the effectiveness of our proposed online algorithm and show that it can save costs significantly compared with always keeping data objects in one tier or always transferring data objects from one tier to the other when their access frequencies begin to vary.
Highlights
In recent years, the paradigm of Storage-as-a-Service (StaaS) cloud has gained more and more momentum
We propose an online algorithm to decide whether to transfer data objects between the cold and hot tiers, and prove theoretically that the proposed online algorithm can achieve guaranteed competitive ratios in saving costs for data objects stored in a two-tier StaaS cloud
In our online cost optimization problem, we need to calculate two breakeven points separately for data objects stored in the cold and hot tiers, which are: (1) for each data object stored in the cold tier, after calculating its break-even point, denoted as βc, when its number of reads accumulated over a period exceeds βc, it will be transferred to the hot tier, otherwise this data object should be remained in the cold tier; (2) for each data object stored in the hot tier, after calculating its break-even point, denoted as βh, when its number of reads accumulated over a period does not reach βh, it will be transferred to the cold tier, otherwise this data object should be remained in the hot tier
Summary
The paradigm of Storage-as-a-Service (StaaS) cloud has gained more and more momentum. It is hard for StaaS cloud users to have a knowledge of the frequency of accesses in advance and it is difficult to predict it For those data objects which are assumed to be seldom accessed, users can choose to store them in cold tiers to achieve lower storage costs. For cloud users, it is difficult to precisely predict such access curves It is important but challenging for a user to decide whether and when to transfer data objects between cold and hot tiers optimally without any prior knowledge of future access frequencies. To address the above issues, in this paper, we propose to use online algorithms to guide StaaS cloud users in making decisions on whether and when to transfer their data objects between cold and hot tiers for achieving cost optimizations, while without requiring any prior knowledge of future access frequencies.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
