Record Setting Hadoop in the Cloud
M.C. Srivas, CTO, MapR Technologies
AbstractWhen MapR was invited to provide Hadoop on Google Compute Engine, we ran a lot of mini tests on the virtualized hardware to figure out how to tune our software. The results of those tests were surprisingly good. The speed and consistent performance provided by Google Compute Engine even at the 99th percentile was very impressive. So we decided to run a TeraSort benchmark just to see how well we could do in a virtualized, shared environment where we didn’t control the hardware, nor control the other non-MapR tenants.