Accelerating the Mobile Cloud Using Amazon Mobile Analytics and K-Means Clustering Matthew Beck, Wei Hao, Alina Campan
Current trends show a move away from desktop computing and toward the rise in popularity of mobile devices. Yet mobile devices suffer from limitations in memory, storage, computational power, and battery life. Many of these limitations can be solved by offloading computations and storage to cloud-based platforms. E-commerce mobile applications designed to serve the global customer base of a retail outlet experience fluctuations in demand for resources based on the location of the users. Given a traditional clientserver architecture, where the server application and database are deployed to a single geographic location, this can cause large disparities in response time perceived by users close to the server location and those at a much further distance. This could cause a loss of business or slow user growth in more distant regions. Using several Amazon Web Services(AWS), this paper tests a proxy system and k-means analysis based data partitioning solution to this issue. The discussion of kmeans database partitioning describes a preprocessing methodology for adapting raw AWS Mobile Analytics log data for use in the k-means algorithm. The paper also compares a few alternatives for distance measurements and centroid computations for use in the k-means algorithm. Experimental results confirm that this approach significantly reduces response time. It also shows that the approach significantly increases server-side throughput.
Mobile Cloud Computing, AWS Mobile Analytics, Proxy Server Caching, K-Means Database Partitioning, E-Commerce Mobile App