The NoSQL Now! conference at the San Jose Convention Center, efficiently hosted by Dataversity, swam with rather new to extremely new data management and analytics offerings, some of which I will write about subsequently. But here are four underlying themes and general observations garnered from the conference:
1. NoSQL is okay with SQL: Several times I heard this sentiment expressed, "NoSQL really means Not Only SQL!" The inference was clear: There weren't many NoSQLers predicting the imminent death of classic RDBMS or SQL. NoSQL opportunities lay in responding to emerging use cases, meeting them optimally and at lower costs. Sometimes NoSQL replaces classic RDBMS hitting the wall, but more often we will find NoSQL working alongside established RDBMS and BI/analytics cousins. ACID was not a dirty word with the majority of NoSQLers.
2. NoSQL Excitement? Yes! NoSQL Confusion? Yes!: The exciting NoSQL supply side offers a dizzying set of diverse solutions, but the associated NoSQL buzz presents a double-edged sword for NoSQL's prospects: On the plus side for NoSQL, potential buyers are drawn to the buzz, curious to find out how to take advantage of all the innovation. On the minus side, those same buyers have a difficult time ferreting through all the technologies and options. A data architect from a large bank I spoke with said, "We have put up our first Hadoop clusters, and are looking what to do next, but I don't know where to look first!"
That said, several database suppliers from the NoSQL movement, such as Couchbase (an open source-based project, related to but not the same as the Apache CouchDB project, for an explanation see here), 10gen (commercial distribution for the mongoDB project) and DataStax (commercial option for Apache Cassandra) are well north of 100 paying customers already, some closing in on 500 paying customers. Even completely fresh-to-America newcomer, Starcounter, an in-memory, NewSQL, scale-in specialist from Sweden, is rapidly moving towards that benchmark 100 number. At this point, the NoSQL community has enough wins to command serious attention from the RDBMS leaders, and indeed Oracle was on hand with its own Oracle NoSQL Database.
3. NoSQL Climbing SHASTA in a V6: In a previous blog post I opined that big data isn't merely about the oft-stated 3 "Vs" (data volume, velocity, variety), but about 6 "Vs", those three plus (4) veracity (i.e. clean, secure data), (5) visualization and (6) business value. While NoSQL seems inexorably linked to big data, NoSQL has its own particular value propositions, which just so happen to spell SHASTA, as in Mount Shasta.
- S = Speed: Nearly every NoSQL database boasted order of magnitude performance gains over classic RDBMS in certain use cases. The speed improvements are typically traceable to (a) data models that fit application cases more readily than classic RDBMS and (b) the use of memory. Then again, there were some unique examples, such as Cray spinout, YarcData, which uses Linux-ized Cray processors to blow through graph analyses at speeds beyond classic RDBMS' imagination - jet plane versus burro.
- HA = High Availability: Distributed database plus cloud and virtualization-friendly architectures, with the aim of ensuring databases are "always available," was touted by a surprising number of NoSQL offerings. You may wonder why databases with "eventual consistency" need to be HA too, but as mentioned above, plenty of these NoSQL databases were indeed ACID-compliant.
- S = Scalability: Perhaps this should be called "scalability without sharding" since "sharding" seemed more of a dirty word than ACID at the event. Regardless, the replication-heavy, schema-light association with NoSQL designs supports scale-out more so than scale-up. However, again, this is the Not Only SQL movement, so there was quite a variety of scaling designs even for NewSQL options.
- T = Total Cost of Ownership: It is pretty easy to claim significantly lower costs, in terms of initial licensing and lifecycle costs, versus the stout margins of classic RDBMS. Basically every NoSQL vendor made that claim albeit without vindication or vitriol.
- A = Access: Almost every NoSQL vendor offers a freebie edition of their database or analytics offering, and nearly every NoSQL database vendor aims to make their database easy to access and use by developers. All kinds of "try before you buy," seeding and imaginative pricing scenarios were evident, all with the same goal: Create a really low barrier of access for potential customers and developers.
4. Objects Redux: OODB stalwarts Objectivity and Versant are finding new veins of ore in the NoSQL gold rush. Objectivity has offered a compelling distributed graph variant called InfiniteGraph commercially for a year now, and Versant focused on "situational awareness" in the context of big data. The bottom line is that a pure object-oriented DB model deals with certain advanced analytics use cases, and mixed transactional and analytical modes, in ways that relational models simply cannot. It was interesting to note that the "object versus relational" arguments were not resurfaced, just the SHASTA arguments for certain use cases.
If you are an IT data architect, or DBA, or data analyst/scientist, what should you infer from the nearly 900 attendees and significant number of vendors displaying their wares at the NoSQL Now! conference? The Not Only SQL movement stands at the tipping point. No longer are the vendors boasting about the number of free downloads, they are boasting about the number of paying customers. For an increasing variety of applications, Not Only SQL options only make sense. Finally, major vendors such as EMC with Greenplum, Teradata Aster, and HP with Vertica, oh, and that little open source Hadoop project through HBase, have already proven the case that databases beyond classic RDBMS work better for certain applications. The Not Only SQL movement is merely underlining, bolding, and adding many exclamation points!...