MapR, based in San Jose, California, provides a commercial version of Hadoop noted for its fast performance. This week at the Strata Conference, I got a chance to talk to the folks at MapR and found out how MapR differentiates itself from other Hadoop offerings.
The fast speed of MapR appears to come from its filesystem design. It’s fully compatible with standard open source Hadoop including Hadoop 2.x and YARN and HBase, but with a more optimized filesystem structure to provide the additional speed boost.
MapR promotes these benefits below.
- No single point of failure
Normally the NameNode is the single point of failure for a Hadoop installation. MapR’s design avoids this issue.
- NFS mount data files
MapR allows you to NFS mount files into an HDFS cluster. This ability saves you time from copying files into MapR and you might not even need tools like Flume. The direct write into the files opens up additional options such as querying Hadoop on near-real-time data.
- Fast access
MapR has clocked the fastest data processing with sorting 1.5 trillion bytes in one minute using its MapR Hadoop software on Google Compute Engine cloud service.
- Binary compatible with Hadoop
MapR is binary compatible with open source Hadoop, which gives more flexibility in adding other third party components or migrating
- Enterprise support
Professional services, enterprise support, and training and certifications
MapR has attracted a number of featured customers including the following:
- Return Path
- Zions Bank
- Live Nation
- Rubicon Project
MapR is also partnering with both Google and Amazon Web Services for cloud-based Hadoop systems.
MapR currently comes in 3 editions.
- M3 Standard Edition
- M5 Enterprise Edition (with “99.999% high availability and self-healing”)
- M7 Enterprise Edition for Hadoop (with fast database)
Additionally, in conjunction with the Strata Conference this week, MapR has announced the release of the MapR Sandbox. Any user can download the MapR Sandbox for free and run a full MapR Hadoop installation within a VMware or Virtualbox virtual machine. This sandbox provides a suitable learning environment for those who want to experience the use and operation of MapR Hadoop without investing a lot of effort in the installation. I haven’t downloaded and installed the MapR Sandbox yet. If you have already done this and tried it out, tell me what you think in the comments below.
MapR website: http://www.mapr.com