A recent infographic from the Data Science Association showed that MongoDB is leading the pack of NoSQL and NewSQL databases in 2014:
I’m not sure exactly where this data comes from, but it matches what I’ve heard anecdotally in the community. MongoDB seems to have a head start for many reasons, including the ease of standing up a new cluster.
Will this trend continue? To begin to answer this question, it’s worth considering the commercial interests behind these databases. This article shows a few metrics on current and projected market share. Of the databases with direct vendor sponsorship, MongoDB leads the pack at 7% compared with 3% each for Cassandra and Riak.
So where does that leave HBase? It’s running a respectable fourth place in the infographic, but that doesn’t tell the whole story. HBase is backed by the entire Hadoop community including contributors from Cloudera, Hortonworks, Facebook, and Intel. Community sponsors for HBase include all of the above plus MapR and Salesforce.com.
Now go back to the market share report and notice that HBase doesn’t have a primary commercial sponsor despite that a significant portion of the market share (and funding) going to companies like Cloudera, Hortonworks, and MapR is backing HBase as well. In the end, HBase may well benefit from being a primarily Apache-backed project that the whole community sponsors and supports, rather than being driven by a single vendor.
This fits into the trend of Apache-backed projects being significantly larger than vendor-backed projects. Apache’s namesake web server is a useful (if inexact) parallel. There are web servers out there that are certainly easier to install and configure, but a huge portion of the world’s websites run on Apache HTTPD. It’s robust, ubiquitous, and has a deep pool of community expertise. The same may be said of HBase in the future – it’s well-supported by every major Hadoop distribution, and it runs on top of Hadoop, yielding some infrastructural savings.
The next couple of years should be very interesting. I’m quite curious if HBase’s Apache heritage will give it the boost it needs to increase adoption in the NoSQL community.