Congratulations to Kamal. He had two papers accepted to CLOUD 2015. “Dynamically controlling node-level parallelism in Hadoop” describes an online mechanism that adapts the node-level parallelism to improve performance up to 28% over best practices. The second paper “Evaluation of MapReduce in a large cluster” discussion our experiences with Pivot’s Analytic Workbench, a 500+ node Hadoop cluster.

The acceptance rate at CLOUD 2015 was 14%, which is 2% for two papers (OK-that isn’t quite how it works).