NoSQL Getting More Popular than RDBMS MPP/Columnar DB’s
IDC’s recent study on Enterprise Hadoop deployments revealed an interesting side fact: NoSQL databases are beginning to motor past RDBMS MPP/columnar databases in usage. In their survey, IDC found that:
Nearly 39% of respondents indicated that they use NoSQL databases like HBase, Cassandra, and MongoDB, while nearly 36% indicated that they use MPP databases like Greenplum and Vertica in conjunction with Hadoop in addition to traditional relational or OLTP databases.
Now I’ve always been a fan of MPP/column databases, and in fact have helped a number of such companies get funding and worked closely with them to get their product out the door. In the right use case situation, MPP/column DB’s can deliver great results.
However, NoSQL databases like Cassandra offer features that you won’t find in these specialized RDBMS engines (e.g. flexible data model, multi-data center/cloud availability zone support, etc.) plus they both overcome some of the I/O limitations found in these DB’s and oftentimes provide the same query firepower that a column solution delivers.
Outside of their high-speed loaders and various asynchronous row-insert-to-column-insert methods, MPP/column DB’s are known for their poor DML performance, which is a trade-off most database pro’s understand. By contrast, a hallmark of a logging-based engine like Cassandra is the incredibly fast insert/update/delete performance you get. This makes Cassandra perfect for high-velocity data use cases like streaming financial updates, sensor data consumption, social media ingestion and so on.
But what is sometimes overlooked with a NoSQL engine like Cassandra is its column-family/wide-row store data model, which can mimic the query behavior you get with RDBMS column DB’s. Using Cassandra’s slice query capability, you can retrieve only the columns for rows you are interested in vs. the entire row. Thus, you can in a sense have your cake and eat it too – fast DML writes with an append-only engine and column query retrieval.
This write/read capability coupled with all the other benefits of a NoSQL model may be part of the reason why NoSQL databases like Cassandra are beginning to move past RDBMS MPP/columnar DB’s in popularity.
If you haven’t given Cassandra and DataStax Enterprise a try yet, why not download it now and kick the tires for yourself? Also don’t forget to take advantage of our new free online training that can help introduce you to developing applications with NoSQL.