Research On Online Cost Optimization Algorithms For Two-tier Cloud Storage Services

Cloud storage has effectively solved the demand of big data storage and access in the current era,and starts to provide services for more and more users.In data management,data temperature is used to represent the frequency of data access.It is usually called hot data that needs to be frequently accessed by computing nodes.Data that is accessed less frequently and does not require real-time access is called cold data.In order to meet different storage service requirements,a new generation of SaaS(Storage as a Service)cloud provides different data storage tiering options.Such As Microsoft Azure cloud providers that are offering hot and cool tiers with differentiated storage cost and access cost to access data.The hot tier has higher storage cost but lower access cost.Thus,making it more suitable for storing data that are frequently accessed In contrast,the cool tier has lower storage cost and higher access cost.Thus,making it more suitable for storing data that are infrequently accessed.Storing data of different access frequencies in the appropriate tiers is a good way to avoid waste of costIn general,users often store the data in the hot tier initially,but as time goes by,the frequency of data access often changes,and the cost of keeping data in the hot tier is higher than that in the cool tier.At this time,the data is transferred to the cool tier,which can make good use of the benefits of the low cost of cool tier.However,because of the costs of data transfer,if the user accesses the data again after the data is transferred to the cool tier,data transfer decision instead bring more cost to the user Therefore,users need to consider the future access frequency of data when making transfer decisions,but it is difficult to accurately predict the future long-term access frequency of data.Therefore,based on the research on the pricing standards and pricing models of the two-tier cloud storage service,we propose an online cost optimization algorithm for the two-tier cloud storage service to help cloud storage users decide when to data objects are transferred between tiers for cost optimization without the user having to know the future access frequency in advance.We theoretically prove that the proposed online algorithm can save the storage cost of data which stored in the two-tier cloud storage.Next,through a large number of experiments,we verified the effectiveness of our proposed online algorithm and showed that with always keeping data objects in one tier,it can significantly save costs.Based on data with a large variation in access frequencies,the decisions made by our proposed online algorithm may be wrong.Therefore,we further propose a storage cost optimization algorithm based on data access frequencies predictions.Experiments show that the algorithm further saves cloud storage costs.