Research And Experiment About The Data Replica Placement Algorithm In Cloud Storage System

With the rapid development of the Internet, the requirements of data storage forvarious types of application are increasing; Internet data is growing exponentially.Cloud storage as a new storage form to provide low cost, high availability data online isdeveloping in recent years. For cloud storage system, you need to ensure highavailability of your data, so that a common approach is to store copies of the data.Data placement algorithm is an important component of cloud storage system.How to ensure high availability of the system, to make the efficiency of data access andto limit the excessive effects of data redundancy at the same time is a matter worthy ofstudy. For now, study of cloud storage system data placement algorithm is still in itsearly stages, there are still many issues that need further study. Decision of reasonablenumber of data copies and the placement policy not only provide good data availability,while improving reading and writing performance of remote data and load balancing,enhance the quality of service of cloud storage services.In response to these questions, this article focus on data placement policy and thenumber of copies decision policy by data management, major work is as follows:(1) Proposed data placement policy in this article is based on the assessmentstaking into account of the value of data read and write, and load balancing, consideringabove factors affecting system performance. Through simulation experiments results ofdifferent factors such as load balancing and consumption of network to analyze theproposed policy.(2) Propose policy based on access trend data to manage data replica number.Propose characteristic function to distinguish hotspot data. After a brief introduction ofsimulation tools, proposed algorithms are applied to experiment environment; theaverage job time and network bandwidth consumption rate are compared with theexisting algorithms.