TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

Relational Technology Has Staying Power

The RDBMS provides transactional safeguards and data management amenities that NoSQL database systems do not now provide -- and probably won't provide for quite some time.

• By Steve Swoyer
• May 17, 2016

Gartner Inc.'s research vice president Merv Adrian is that rarest of industry analysts: someone whose personality can make even the driest news into interesting content.

Adrian, an industry veteran, was at his most interesting recently. The occasion was a sneak peek at some of Gartner's DBMS market research data. The context was Adrian's Gartner blog, where he often writes with a frankness that probably keeps his Gartner PR handlers up at night. This time Adrian's aim was to put the growth of NoSQL into context.

Specifically, he addressed two things: first, the extent to which NoSQL DBMSs have grown their combined share of overall DBMS market revenues; second, whether this growth -- if measurable -- had come at the expense of the entrenched RDBMS platforms.

The short answer to both questions is "yes." NoSQL revenues are growing. Between 2014 and 2015, for example, revenues for the Hadoop platform players (Cloudera Inc., Hortonworks Inc., and MapR Technologies Inc.) basically doubled. Based on Gartner's market research data, it seems that NoSQL -- a category that for our purposes includes Hadoop -- has stolen share from the venerable relational database.

The thing is, RDBMS revenues are growing, too. Not nearly as fast as those of NoSQL systems, but steadily.

The upshot is that the DBMS market is healthy -- very, very healthy. Everybody's benefiting.

Adrian's blog provides some much-needed context. In the last half-decade, the DBMS market leaders -- all of them purveyors of relational database management systems -- have lost about 2.4 percentage points of market share, dropping from 91.4 percent of the market in 2011 to 89 percent in 2015.

At the same time, however, the DBMS market as a whole has grown -- 8.9 percent in 2014, 8.7 percent in 2015. Fundamentally, what we're talking about is how to interpret this growth. Almost everybody's growing. Some, starting from lower baselines, are just growing much faster than others.

In a market valued at $35.9 billion, the combined revenues of the top five NoSQL vendors and the three biggest Hadoop vendors amount to slightly less than $700 million, Adrian said. That's a relatively small slice of a pretty big pie.

In other words, even though we might speak of NoSQL "eating" into the share of the dominant RDBMS vendors, this is to some extent an interpretation of the facts. Again, it isn't as if the RDBMS market has been shrinking.

Sure, Oracle Corp. "lost" 1.5 percentage points of market share between 2011 and 2015, dropping its share of the DBMS market down to 41.6 percent. That still works out to almost $14.6 billion.

At the same time, Oracle has managed to grow its revenues year over year. This being Oracle, there's no lack of cynical explanations as to why this is the case. Last December, however, Db-engines.com, a site that uses a mix of criteria to rate the popularity of basically all extant database systems, proclaimed Oracle its "DBMS of the Year," citing that platform's strong, consistent growth.

Similarly, Gartner says Microsoft Corp. actually grew its share (by 0.8 percentage points) of the DBMS market between 2011 and 2015. Redmond now accounts for $7 billion of the overall DBMS pie.

This isn't a "circle the wagons, the RDBMS is just fine" piece. It's a plea for perspective.

The RDBMS addresses a number of critical needs that are not -- or cannot easily be -- addressed by upstart technologies. To cite just one example, very few of the available NoSQL DBMSs -- including banner-leaders MongoDB, Cassandra, Couchbase, and vanilla Hadoop -- support ACID-level transactional guarantees.

For OLTP applications, ACID isn't an abstraction. It isn't overkill. It's a sine qua non requirement. DB2, SQL Server, MySQL, Oracle 12c, and Sybase Adaptive Server Enterprise (ASE) -- to cite just a few examples -- are ACID-compliant RDBMSs.

Google Inc.'s F1 -- a SQL layer for its "NewSQL" Spanner distributed database -- implements ACID-compliant transactional safeguards, too. Several commercial or open source add-on technologies for Hadoop do as well (Actian's Vector, Dell Greenplum's Hawq, for example).

Organizations need relational databases. Do they need them less than they used to? That depends. In the past, the RDBMS was misused to support applications for which it was not designed and is not ideal. There was no reason -- none at all -- to use an RDBMS to support internal or departmental Web app dev projects, yet organizations did just this. A NoSQL store such as MongoDB was and is a far better option. The popularity of NoSQL means organizations are going with that option.

To be sure, it also means they're tapping NoSQL systems to address use cases for which they aren't designed and aren't ideal. However, in spite of the fervent desires of enterprise developers -- for whom the RDBMS and its object-relational mapping techniques are a perennial source of (not unreasonable) frustration -- the RDBMS isn't going anywhere. Organizations need relational databases to provide the transactional safeguards and data management amenities that NoSQL DBMSs do not now provide -- and probably won't provide for quite some time.