Cloud computing systems provide scalable infrastructure to store and process Big Data generated by various organizations. Distributed file system (DFS) is used as the main storage element in a cloud computing system for storing and accessing Big Data. Improving the performance of the read operations in the DFS is one of the important research issues as more frequently the users perform read operations on the DFS and less frequently the write operations. In this paper, we have developed two novel read algorithms for improving the performance of the read operations of the DFS by considering the presence of the client-side caches, global cache and speculative processing. The main advantages of our algorithms are (i) Reduction in read access time (ii) Write operations do not perform caching and do not require the execution of cache invalidation or synchronization protocols. Our performance evaluation results indicate that the proposed algorithms perform better than the algorithm which does not use caching and speculative processing. Large amount of data is getting generated due to the extensive use of web applications by billions of users around the globe. The organizations which has deployed web applications are pondering over solutions for scalable storage and faster access of large data. Distributed file systems (DFSs) have been emerged as efficient storage solutions so that the data can be stored and accessed efficiently. Modern cloud-based data centers have been using the DFS as main storage component. Improving the performance of read operations have become important as most of users of the web applications carry out read operations in the web. Caching and speculation-based approaches are proposed in the literature for improving the performance of read operations in the DFS. In this paper, we have proposed an anticipated parallel processing-based algorithm which considers the presence of multi level caches namely local cache, local cache of neighbouring node (nearby cache) and global cache. We have carried out the performance evaluation of the algorithms using mathematical analysis and simulation modeling. The results of the analysis indicate that the proposed algorithm performs better than the speculation-based algorithm proposed in the
Volume 11 | 01-Special Issue
Pages: 1667-1673