Performance vs cost for windows and linux platforms in Windows Azure cloud

Abstract

The cloud resources can be efficiently utilized using different parallelization methods and techniques. However, most of them depend on the operating system and its runtime environment. In this paper we perform series of experiments to analyze the performance of dense matrix-matrix multiplication algorithm on the same hardware infrastructure using parallel threads on different platforms in Windows Azure, i.e., different operating systems and runtime environments. Contrary to the hypothesis that Linux platform provides better performance, the results of the experiments show that Windows platform runs up to 3.01 times better, especially for problem size that can be placed in last level cache and will not generate a lot of cache misses. We also determine the regions where a particular operating system is a better solution if cost or performance are considered.