Enhanced replication strategy with balanced quorum technique and data center selection method in cloud environment

Digital data are growing tremendously, and the massive amount of data are core contributors to data evolutions across the globe. In this Industry Revolution (IR) 4.0 era, heterogenous data are generated in various sources and platforms. In order to store these huge volumes with heterogenous data categories, cloud computing became the mainstream solution to provide multiple services to keep safe, process and distribute the data. As each data is substantial to everyone, cloud computing emphasises and facilitates a fast and flexible platform to users. Despite continually standing as a resilient storage provider to heterogenous data, the prominent issue encountered as performance challenges is to accommodate sufficient services with high data cloud storage. The issues are crucial because poor data availability and accessibility are often influencing delays on data retrieval which ultimately leads to system performance degradations. In order to mitigate the issues, ‘cloud data replication’ is commonly implemented for better data performance and promising business continuity. Cloud replication is recognised as storing more than one copy of data in multiple distributed storage nodes. Additionally, cloud replication is well-known as a comprehensive technique which is capable in serving high data availability, faster response time, and better fault tolerance yet cost-effective for both cloud users and provider. Regardless of the replication valuable keys, occasionally this technique has the tendency to attain performance degradation issues too. Some existing research works did not considers substantial factors in determining popular files. As a result, the poor popular file selection strategies affect file download and lead to ineffective response time. Besides, there are also issues that arise during data placement techniques consuming extensive space and high replication time due to replica copies located in every storage node in the same cluster environment to achieve high data availability. Furthermore, there are also problems identified when choosing an appropriate data center to store replica copies caused by ineffective data center selection criteria. Holistically, there are many previous studies which have enlightened issues in cloud replications such as delayed response time, high replication time, extensive storage space, and massive network consumption due to inefficient replication strategies implemented in cloud environments. Therefore, to evade the negative probabilities and satisfy users, an established and extensive cloud replication strategy is significant to be deployed in a cloud environment to enhance the overall replication performance. In this research, there are three (3) main contributions proposed for a cloud replication environment. Firstly, this research proposed an Enhanced Replication Strategy (ERS) that improves response time for file download in a replication cloud environment without neglecting the popular file selection. Secondly, this research proposed a dynamic Balanced Quorum (BQ) technique for replica placement to reduce storage consumption without disregarding a faster replication time and preserve data availability. As for the third contribution, an efficient Data Center Selection Method (DCSM) was proposed to ensure files are available in the best local data center; thus, this research is able to facilitate efficient network usage and minimise replication frequency in a cloud replication environment. The proposed ERS, BQ and DCSM are agile depending on the replication environment requirements and dynamically adapt user access patterns, guarantee that essential data is always available and accessible by users. Thorough experiments were conducted using the ‘CloudSim’ simulation tool. The dataset generated by CloudSim was in random structured master files in various sizes. The simulation results were analysed and the analytical graphs are presented in the discussions. In order to validate the competence of these proposed algorithms, results were compared with another similar research work known as DPRS algorithm. Apparently, the proposed ERS, BQ and DCSM in this research outperformed the DPRS, concurrently evidenced 10.47% improvement in average response time, reduced storage consumptions are 11.43% and 31%, accelerated replication time with 6% and 20%, sustained high data availability between 98.59% to 99.03%, offered effective network usage reduced by 20% more efficiency and finally betterments on overall replication frequency with 14%. Individual measurement metrics in this research was outpaced the other existing method, the DPRS and contributed better performance in cloud replication environment.

Text FSKTM 2022 13 IR.pdf Download (1MB)

Actions (login required)

View Item