Continuing a look at the very hot Hadoop space, today I wanted to spotlight a new company that has made a big bang since coming out of stealth back in June of this year. With just one quarter under their belt, MapR has over 60m of deals in their pipeline, a major OEM deal with EMC and perhaps the best Hadoop distro out there. How and why has this San Jose based start up done it? I spoke with Jack Norris, VP of Marketing at MapR to get a better idea.
MapR has a management team with deep experience in storage, open source and virtualization. They raised venture funding from NEA, LightSpeed Venture Partners and Redpoint Ventures. A big piece of the Hadoop equation which they make better is that they offer loopless storage services. Improving the native Hadoop HDFS, the MapR solution is a significant upgrade. Some of the limitations of HDFS according to Norris are HDFS works like writing to a CD-ROM, it is written in Java and uses the underlying Linux file system. The MapR system improves on each of these.
There is more to MapR though then just an HDFS replacement. MapR has its own distributions of Hadoop. The advantages of the MapR distro are summed by the following:
In terms of dependability, here is what MapR offers:
The advantages to the MapR Hadoop version were compelling enough to have EMC choose it as the Hadoop version they are selling. A major partner like that allows MapR to compete in a market where companies like IBM distribute the Apache version (until recently IBM was distributing a custom IBM version) and companies like Cloudera and others are also pushing Hadoop distributions.
MapR offers two versions of their Hadoop distro. One is a free community version called M3. They then offer a commercial version called M5. It should be noted that while M3 is free, it is unclear if it is open source or not. I am checking on that now. The differences between the two versions are:
So it looks like MapR is going to be a power in the Hadoop space. To make this much progress so quickly speaks volumes not only to MapR, but to the Hadoop market itself.