Introduction
Cloud data storage became popular because handling huge amounts of data is required & the size of data is increasing day by day. In cloud data storage systems, replication is used for creating similar copies of some data items stored in the cloud. Latency can be minimized after creating replica at various data centres from source data centres. Response time is also an important factor these days. So replica creation & its placement is important for the system performance & decreases the latency of the application deployed over the cloud.
Every user wants to use only that replica that is geographically nearby its location so response time can be minimized. For continuous data, a system must be designed with no failure which is practically impossible. We can create a replica of data so that if some of the replica is not able to serve the request another replica can process the request & application execution completes successfully. We cannot place replica at all the nodes because every node has some limitations, some of them have limited storage & some of them can be minimum bandwidth so we must find out a new technique for replica placement.
Issues Related to the Replication and Possible Solution
The two major problems which are minimized by data replication in cloud storage system is data availability & performance of the application. For a file, how many replica should be created is the first issue in this type of situation & then where to place those replicas is the second challenge.
Which files should be replicated, this can be obtained by the analysis of the user request.
Replication can be categorized in two ways
- static replication
- dynamic replication.
In static replication, which data is replicated at which data centre is decided in advance but in case of dynamic data replication technique no. of replica & placement of replica is decided according to the situation of the running application. In dynamic replication technique replica is created & placed at an appropriate place according to the demand frequency of data items.
In my suggestion using a multi-criteria decision making method can be useful for replica placement and we can consider the following criteria such as user request, storage, bandwidth, system performance etc for taking decision about placing the replica.