HAMR has been designed from the ground up to achieve the ultimate level of performance. We pitted HAMR head-to-head with other market leading products, and they didn't stand a chance. Don't take our word for it, see for yourself!
Our engineers cooked up a Naïve Bayes trainer algorithm running in HAMR, then compared it to a standard Hadoop 2.5 installation running Mahout 0.7. The result? A whopping 87x performance gain on the largest dataset.
PageRank is a common link analysis algorithm, so why not implement that in HAMR and see how it stacks up? We ran our benchmark and compared it to a Hadoop 2.5 and a Spark 1.0.1 implementation. Guess what happened?
How would the competition handle an algorithm that generated an explosive amount of intermediate data? K-Cliques is one such algorithm, so we gave it a shot. Spark kept dying with 'out of memory' errors. Hadoop managed to run it, but took its good old time. HAMR breezed right through it without a care in the world.