Dynamic data replication and placement strategy in geographically distributed data centers

With the evolution of geographically distributed data centers in Cloud Computing landscape along with amount being processed these centers, which is growing at an exponential rate, processing massive applications become important topic. Since a given task may require many datasets for its execution and are spread over several different finding efficient way to manage storage across nodes system difficult problem. In fact, time might be influenced by cost transfers, mainly depends on two criterias. The first one initial placement input during build-time phase, while second replication runtime phase. explicitly considered when migrated order make them locally available wherever needed. Data challenges Computing. Nevertheless, studies focus or exclusively. this paper, combination strategy followed dynamic management proposed, purpose reducing associated all transfers between (distant) centers. Our proposed approach considers main characteristics center such as capacity read/write speeds efficiently store datasets, our three parameters: number replicas system, dependency tasks decision whether keep delete determined fulfillment those parameters. estimates total well monetary cost, considering activity. experiments conducted using Cloudsim simulator. obtained results show that strategies produce overheads compared both without (by 76%) selected from Kouidri et al. 52%), improving financial cost.