As Web-scale NoSQL technology looks to find a deeper footing in the enterprise, there may be as many stumbles as steps forward. That was an underlying theme at the NoSQL Now 2015 conference, where issues with service-level agreements (SLAs), data analytics challenges and a lack of skills were often part of the discussion.
Of course, NoSQL has had victories. Modern NoSQL databases were forged in part by DevOps advocates at Web successes like Amazon, Google and Facebook. These businesses were ready to embrace whatever solved their problems as online operations scaled up drastically and conventional SQL relational databases with rigid schemas fell short in meeting their processing needs.
The developers thrived on the wings of an open source software ethos, and their bosses generally let their inventions go open source, too -- or at least let them publicly share the details of what they were doing. Sometimes working with such blueprints, and sometimes without, a host of startup vendors -- including the likes of DataStax, MongoDB, Riak, Couchbase, and Neo Technology -- have helped define the nascent NoSQL market, and newcomers are still emerging.
Fast development and performance that can scale massively have been NoSQL's calling cards; activity tracking, aggregation pipelines and high-volume transaction applications have been among the sweet spots for business uses. At NoSQL Now, held last month in San Jose, Calif., it seemed apparent that graph databases -- still something of an outlier among the various categories of NoSQL software -- were finding wider use, in particular a "triple store" variation on graph technology for use in applications that bring semantic order to disparate pools of data.
But obstacles lie in wait on the way to NoSQL enterprise nirvana. For starters, the skills needed to really get high performance out of NoSQL systems aren't always readily available to organizations. The skills required to keep such systems running steadily are also in short supply, making some NoSQL SLAs dicey at best. And the schema flexibility that helps speed NoSQL development can cause backups on the back end when business-side folks look to create reports based on NoSQL applications.
Looking to bridge some NoSQL gaps
Some of the obstacles are cultural. There have been ways of doing things in the relational database world that will either have to be adapted or jettisoned as NoSQL technology settles in alongside its predecessor. For example, conference speaker Dan Sullivan said good support for building NoSQL data models is basically missing at this point, although people are working to fill that gap.
"With relational modeling, you look at the data, and you model based on the relationships between the data," he said. "In NoSQL, you look at your queries, and then you design according to that." But the queries being run by users "are constantly changing," said Sullivan, who is an independent database consultant and author of the recent and worthwhile NoSQL for Mere Mortals. As a result, he added, "we have to realize that now our models are implicit, and that new queries may require new data structures."
Ravi Krishnappa wants to be positive about using NoSQL technologies, but he too sees barriers. "I'm not too optimistic that these NoSQL databases are enterprise-ready," said Krishnappa, a senior solutions architect at data storage vendor NetApp Inc. To him, enterprise-ready means durability -- and that is a trait many of the NoSQL database vendors are still working to achieve.
There's more. Familiar RDBMS scaffolding and controls still aren't there with the NoSQL platforms, in Krishnappa's estimation. He lists such missing pieces as design menu systems, security and reporting tools. That last feature has become an essential part of SQL over the years, but it's an absent or hard-to- achieve element in most NoSQL implementations today.
Development teams can fill in where pieces are missing -- vendors, too. But other conference participants agreed with Krishnappa's contention that "You don't find too many NoSQL skills in the marketplace." That can be a gating factor, at least for companies without the deep pockets and workplace appeal of Google and Facebook.
Fork ahead in the NoSQL data road
Looking forward, what we may see is a branching, where raw, original-style NoSQL serves the needs of pedal-to-the-metal developers, and something else evolves to meet the stricter needs of enterprise shops. That view came out in a conversation with Adam Wray, CEO of NoSQL database provider Basho Technologies Inc.
"We see a trend in the market where there are two basic approaches," said Wray, whose company offers a commercial version of the Riak NoSQL platform. First, there's open source NoSQL that your engineers wrench at, and you build it all yourself, he said, noting that this isn't necessarily a bad approach in some cases. "Then, there's the enterprise point of view," Wray said. "In the enterprise, they have other ideas about how people should spend their time."
It isn't always a question of enterprise data developers' abilities, Wray continued. It's more a question of whether they should busy themselves with building features for high availability, configurability, manageability, scalability and some other abilities to be named later into NoSQL systems.
All of this shouldn't obscure the many uses that NoSQL technology is finding in the enterprise and elsewhere. It should tell us, however, that there is still trial-and-error in the offing as NoSQL data architectures find their fit.
View a video on the latest MongoDB database doings
Read about a version update to the Couchbase NoSQL server