Cost Reductions for Hadoop Clusters Using 10GbE Rather Than 1GbE Network Adapters

Emulex Corporation announced the results of a Hadoop performance study at Cloud Connect comparing the performance of 10GbE versus 1GbE for Hadoop cluster environments.

The results of the study conducted by the Emulex Advanced Development Organization demonstrate how Emulex OneConnect 10GbE Network Adapters quadruple the performance over 1GbE networks for Hadoop cluster environments, addressing the ‘Big Data’ performance needs of cloud providers and users.

Historically, Hadoop has been designed to run on commodity servers and is optimized for 1GbE networks. However, today’s commodity servers are delivering a substantial leap in processing and I/O performance. With recent and ongoing advancements in processor and storage technology, the legacy 1GbE network has become the bottleneck in Hadoop clusters.

The study reveals that leveraging Emulex OCe11102 10GbE Network Adapters quadruples the transfer rates for Hadoop clusters, increasing network throughput, resulting in faster data ingestion and replication. As a result, more CPU cycles become available for data analysis, improving system response time and reducing the required number of Hadoop cluster nodes. Cluster size reduction and increased transaction response time result in significant cost reductions in implementing, operating, and maintaining a Hadoop cluster.

To read this study (you need to register)