Reseach On Data Placement Strategy For Data-intensive Applications In Cloud

Recently, with the continuous development of Internet and cloud computing, data-intensiveapplications have been received more and more attention. These applications will produce a greatquantity of data so that facing some huge challenges for establishing cloud computing system. Inorder to achieve rational data placement, the article analyzes the related concepts and keytechnologies of cloud computing deeply, then the article proposes one data placement strategyusing improved K-means clustering algorithm and another data placement strategy using improvedK-means clustering algorithm based on Fisher’s linear discriminant analysis, which purposes are toincrease the number of system’s data movements.The selection of initial clustering centers has been optimized using improved K-meansclustering algorithm, which improves the quality of clustering results. According to the improvedalgorithm, the article designs corresponding data placement strategy immediately. The results showthat the improved algorithm decreases43%of the number in data movements than the randomalgorithm. Another data placement strategy using improved K-means clustering algorithm based onFisher’s linear discriminant analysis which combines Fisher’s linear discriminant analysis withimproved K-means clustering algorithm, which objective is to refine the boundary of data centers.According to the improved algorithm, the article designs corresponding data placement strategyimmediately. The results show that the improved algorithm decreases26.6%of the number in datamovements than the random algorithm.Using one data placement strategy using improved K-means clustering algorithm and anotherdata placement strategy using improved K-means clustering algorithm based on Fisher’s lineardiscriminant analysis can reduce the data-intensive applications’ number of data movements acrossmultiple data centers in the runtime and effectively improve the clustering quality and the overallutility.