Relational database leader Oracle Corp. is so intricately associated with the SQL-based approach to databases that...
its NoSQL efforts are bound to be overshadowed. When enthusiastic NoSQL upstarts launch jabs at the Redwood, Calif.-based SQL behemoth, they usually don't mention Oracle's NoSQL lineage, although, in many cases, it predates their own.
Oracle's own sales teams may not mention the Oracle NoSQL offering either. For these and other reasons, Oracle NoSQL may seem like an oxymoron. This could change as more users try to quickly field distributed Web and cloud applications that challenge traditional SQL's hold on business software.
Oracle's NoSQL story tracks back at least to 2006 when it purchased Sleepycat Software, a company formed to commercialize the open source Berkeley DB key-value data store. While Oracle has added to the original Sleepycat software, and contributed to open source Berkeley DB, it is a relatively small part of the company's data efforts.
This ''cat'' may not be sleepy much longer. The company is preparing to update Oracle NoSQL next week with enhancements that speed data ingestion, while improving search and other traits. The capabilities are intended to better position the product in emerging high-performance applications where NoSQL is increasingly seen as a natural fit.
"In the last few years, what we have seen is that customers need more than just relational technology to meet their data management needs," said Ashok Joshi, senior director of NoSQL, Berkeley Database and Database Mobile Server at Oracle. Applications apt for NoSQL deployment include sensor data capture, customer call records and fraud detection, among others, according to Joshi.
Letting 'Sleepycats' lie?
A generation of new apps puts some pressure on Oracle to focus effort on its NoSQL entry. The new software has made a mark on the data landscape, according to industry analyst Curt Monash.
"There is no doubt that NoSQL has captured, and deservingly so, a non-trivial share of the market of new database applications," said the president of Monash Research, indicating that a key benefit of new NoSQL applications is that they are being built without the big upfront schema design and model build-out that over the years has become associated with SQL.
Monash suggested Oracle has more work to do to catch up with some technologies now offered by the much smaller NoSQL crew.
"There's little reason to choose Oracle NoSQL over independent alternatives," Monash said, "unless you happen to have a deep and abiding love for Oracle as a company."
Matt Aslett, a 451 Group analyst, said Oracle's original interest in Berkeley DB and Sleepycat seemed to be based on its potential as an embedded database, but that the company also was pleased to have another key-value data store in its portfolio.
Since acquiring Sleepycat, the product has been maintained and updated, he noted, "but it has never been a big focus for them." He said Oracle NoSQL updates are in part "a recognition that there is a requirement for such high-performance databases." Still he sees Oracle NoSQL's major fit as a part of a larger portfolio of Oracle database software, rather than a standalone entry.
"I have not come across it except as a complement to Oracle's Big Data Appliance or its Hadoop [products]," he said.
Get your NoSQL program here
Should Oracle take on the NoSQL market wholeheartedly, it will find something of a moving target.
NoSQL database makers are looking to improve their readiness for enterprise application deployment, and to take on features beyond the original fit-for-purpose designs that gained them their first reputation. The NoSQL market is still wide, and product updates frequent. NoSQL players with updates in recent weeks include:
- Aerospike Inc., which disclosed work on geospatial features for its database server;
- Basho Technologies Inc., which added an open source version of its Riak time series database;
- Cassandra specialist DataStax Inc., which showed what it described as an experimental version of its upcoming DataStax Enterprise Graph system;
- MarkLogic Corp., one of NoSQL's earliest advocates, which previewed MarkLogic 9, including an API that lets developers create queries that combine documents, triples, and row data; and
- Neo Technology, which rewrote its data engine to boost graph database scalability and added a new binary wire protocol, also known as Bolt to its Neo4j 3.0 product.
For its part, Oracle is looking to enhance Oracle NoSQL with new enterprise and performance traits. Joshi points to age out of expired data and predicate pushdown as respective examples.
The Oracle age-out facility can retire data that spans time frames beyond those of interest to users. He said that, in sensor data collection particularly, analysts may have interest in only three to six months of data. But, he added, they can save it all if they want to.
He said predicate pushdown improves performance by moving some SQL query work into the NoSQL database. Using the technique, "you only need to send results back to the Oracle SQL database for further analytical processing," he said.
"Some data in Oracle SQL" and "some data in Oracle NoSQL" are recurring themes. "A common scenario we see is using the two together," Joshi said. Clearly, keeping long-standing SQL customers from going off the ranch to new NoSQL alternatives is a driver at the company.
Read about ways to future-proof your database plans.
Learn about NoSQL ins and outs.
Take a look under the hood of a NoSQL database.
Read about how to tackle NoSQL performance management