Ezakus, a leading data management platform, relies on Hadoop to process 600 million digital touch points raised by 40 million users and mobile users.
Fast growth created challenges in managing Ezakus’s existing Hadoop installation, so they tested different alternatives for running Hadoop. Their benchmarks found that Hadoop on Google Compute Engine provided processing speed that was three to four times better than the next-best cloud provider.
“Our benchmark tests used the Cloudera Hadoop distribution”, said Olivier Gardinetti, CTO. “We were careful to use identical infrastructure - the same logical CPU count, the same mem capacity and so forth. We also ran each test several times to ensure that outliers weren't skewing the results.”
When using MapReduce for basic stats processing of 20,469,283 entries along their browsing history over 1 month, Compute Engine computed the stats in 1 minute and 3 seconds, four times faster than the alternative tested. When more complex queries were run in a second test, Compute Engine computed in 7 minutes and 47 seconds, 3 times faster than the closest alternative which ran at 23 minutes and 31 seconds.
Ezakus can now provide more performance and predictions and serve more clients, “because we can more easily deploy all the servers in a very short time,” said Gardinetti. To learn more about their migration to Google Cloud Platform and subsequent results for their business, read the case study here.
-Posted by Ori Weinroth, Product Marketing ManagerImage may be NSFW.
Clik here to view.
The post Ezakus runs 1000 nodes of Hadoop on Google Compute Engine appeared first on Platform as a Service Magazine.
