MapR M7 release offers alternative architecture for Hadoop's NoSQL database to deliver better reliability, performance, and manageability.

Doug Henschen, Executive Editor, Enterprise Apps

October 23, 2012

3 Min Read

There are now scores of NoSQL products on the market, but none of them have the adoption potential of Apache HBase, the NoSQL database native to the Apache Hadoop platform.

Unfortunately, HBase is also an immature platform that suffers from growing-pain flaws including instability and cumbersome administration. Enter MapR, which announced Tuesday an M7 software distribution aimed at delivering high-performance and easy-to-administer Hadoop and HBase in one deployment. In the case of HBase, MapR says it delivers at least two times faster performance than HBase running on a standard Hadoop architecture.

Because it runs on Hadoop, HBase promises both scalability and the economy of sharing the same infrastructure as the world's most popular big data processing platform. HBase offers higher scalability than NoSQL products such as MongoDB, according to MapR, and it delivers better database consistency than products such as Cassandra. Despite its flaws, HBase is already used by nearly half of Hadoop users, according to MapR.

[ Want to learn about MapR cloud services options? Read MapR's Google Deal Marks Second Big Data Cloud Win. ]

To address HBase weaknesses, the M7 release dispenses with infrastructure and administrative steps that the vendor deems unnecessary and an obstacle to performance. This follows the pattern that MapR set with its previously available MapR M5 distribution, which replaces the open source Hadoop Distributed File System (HDFS) with a proprietary file system that MapR touts as better performing and more reliable than HDFS.

More specifically, M7 does away with region servers, table splits and merges, and data compaction steps, for example, and it implements a simplified architecture exclusive to MapR for snapshotting, high availability, and system recovery.

"We support instant recovery on MapR clusters because there's no separate region-server process, and the HBase tables read directly from the snapshots," MapR VP of marketing Jack Norris told InformationWeek. "You get consistent, low-latency performance on HBase, and that expands the use cases that are viable for running NoSQL on Hadoop."

M7 creates a unified data layer that makes the same snapshots, mirroring, namespace, files, and tables accessible from the same management regime. This simplifies data access and administration and improves performance for both Hadoop and HBase on the same, shared infrastructure, according to Norris. "A lot of companies have implemented separate Hadoop and HBase clusters to avoid some problems that we're eliminating."

With M7, developers can use MapReduce functions or create tables in HBase at will without limits on the number of tables you can implement on any one cluster, he said, noting that standard HBase deployments are limited to 200 tables per cluster.

For MapR customers running the vendor's M3 or M5 distributions, upgrading to the M7 implementation of HBase will involve a standard table-copy procedure to move the data into the new database, according to Norris. The M5 and M7 releases include snapshotting, mirroring, and data-protection features as well as enterprise support that is not available with the M3 release.

As for customers who aren't using MapR, the question is whether the vendor's proprietary approach will be a stumbling block. MapR competitors including Cloudera and Hortonworks are quick to point out that MapR's M5 distribution relies on non-open-source software. Their spin is that that turns a customer's Hadoop deployment into a non-standard environment. MapR's spin is that it's delivering better manageability and performance while exploiting the best that Hadoop has to offer.

"The concern that people have when they hear 'proprietary' is about being locked into some vendor's format and locked into where the data resides," Norris said. "But when we've done these innovations, we've actually opened up the data so it's much easier to get the data in and out of Hadoop and, now, HBase."

About the Author(s)

Doug Henschen

Executive Editor, Enterprise Apps

Doug Henschen is Executive Editor of InformationWeek, where he covers the intersection of enterprise applications with information management, business intelligence, big data and analytics. He previously served as editor in chief of Intelligent Enterprise, editor in chief of Transform Magazine, and Executive Editor at DM News. He has covered IT and data-driven marketing for more than 15 years.

Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.

You May Also Like


More Insights