Joseph Issa
performance analysis, cloud computing
Given the exponential growth in cloud computing, it becomes important to evaluate and characterize the performance of a cloud cluster and understand the hardware and software bottlenecks that affect performance. It is important to analyze the underlying hardware in a cloud cluster by enabling or disabling certain hardware features to achieve the maximum performance possible. In this paper we present a performance evaluation and analysis for Hadoop Kmeans workload which is a compute-bound workload in iteration phase. We also propose a performance estimation model that can predict the performance for Hadoop Kmeans by modeling different processor micro-architecture parameters. The model is verified to predict performance with less than 10% error margin relative to a measured baseline.
Important Links:
Go Back
