March 2, 2013

On the Couch with a NoSQL CEO

Nicole Hemsoth

Couchbase, which has built a business around the self-titled open source NoSQL database, has a clear view of the changing database landscape and sees some crumbling monoliths on the horizon.

This week at the Strata Conference in Santa Clara, we sat down with the Mountain View-based private company’s CEO, Bob Wiederhold, to talk about the NoSQL ecosystem in general—and Couchbase’s expanding role in it.

For those who followed the momentum around Membase and CouchDB a couple of years ago, watching what’s happening with Couchbase Server and the entire NoSQL ecosystem is confirmation that the belief that some “niche” database approaches wasn’t unfounded. Couchbase is now powering household name brands, including Orbitz, Thompson Reuters and the ad platforms for AOL, among others.

We talked briefly with the CEO about how the NoSQL race is shaking out its winners in critical areas. Naturally, Wiederhold believes that in the key value realm, Couchbase is leading the pack. However, he notes that while document database-driven companies like MongoDB and 10gen have remained in heated competition, a pairing of approaches is what is really going to sing for performance and scalability-aware users.

He argues that there are three areas of consideration for developers in a post-relational world–scalability, performance and ease of development. While Wiederhold says some of their more recent work on the ease of development side have brought them up to speed, where key value has always shined is on the performance and scalability side.

Wiederhold claims that one attractive element Mongo has is that it has really focused is on creating ease of development with a very rich developer environment and lots of features to do indexing and querying. This has been the key to their success, he notes, especially those working on relatively small applications that are never going to run on more than one server. Here, he claims, you don’t need scalability, and oftentimes those simple applications don’t need very high performance, so, as he puts it, “they’ve gotten very popular and very successful largely based on this ease of development.”

But key value comes with that same built-in ease, he claims, since it revolves around storing blobs of information associated with the key names until it’s called upon to spit back the associated blob of information.  “So for the developers that really went with key value, what they were really looking for was scalability and performance,” Wiederhold says. “ To be able to very easily scale from 2 node to 10 nodes to 20 nodes to 100 nodes, and be able to do that on the fly, dynamically etc. “

The NoSQL CEO notes that this is one reason why they’ve been particularly successful with advertising platforms. These are a prime fit for key value databases because, as he described, the need for ultra-high performance and scaling is critical. For instance, when you come to a website, the ad platform has 40 to 50 milliseconds to paint an ad on the screen from the moment your cookie identifies you. On a platform like AOL, that cookie needs to be streamed against 750 million profiles to get access to that profile of your likes and dislikes, and then pulling some heavy logic to figure out “ok, I’ve got 4000 advertisers, and are you 25-35 and a woman and you like clothes, and so I’ll paint this advertisement on your screen.” Wiederhold argues that for these users, there’s little need for a bunch of weighty developer features–they just want massive scalability and ultra high performance.

In other words, the key to the key value lock is to be really strong in scalability and performance, but he openly admits that they are not the only ones who stand out here.  MongoDB also offers solid scaling and performance, he said, adding that “ Now that we also have ease of development, we think we’re a great alternative to MongoDB, particularly if you have higher performance requirements or if you need to scale even to 5 or 6 node – 20 nodes, 100 nodes etc.”

While others in the column and graph database camps (Cassandra and Neo4j, as examples, respectively) are also making inroads on the all-important scalability and performance fronts, he notes that those are still somewhat specialized and not as simple to deploy and use (Cassandra being the real target of this statement). The real power behind a NoSQL database offering, he suggests, is to begin taking the best from all database worlds and building out offerings that meet actual use case requirements—especially for media, gaming and other web-based business operations.

These needs are pushing their drive to extend their current offerings to bring them in closer heat with Mongo, 10gen and others. For instance, this week at Strata, Couchbase fleshed out their Couchbase Server 2.0 release by adding another open source piece to its ranks via a plugin for the open source Elasticsearch functionalities. This bolstered the new document capabilities that were added in December by letting developers incorporate search, relevance-based queries and more comprehensive analytics on its platform—all very useful features for those building applications that gather third party data from the massive wells of Twitter and Facebook, for example.

Next — The Giants are Falling>

 

At the end of last year, the open source project spearheaded the newest 2.0 release of its main claim to NoSQL fame, Couchbase Server.  The goal was to extend the key value database capabilities by adding a new set of document database possibilities, thus bringing it into a new level of competition with document database company, MongoDB.

As Widerhold noted, “We think that key value and document are going to combine into one – a document database is really just a specific type of key value database – so you’re just storing JSON documents into value and doing queries.” This extension into the document realm is critical for the company’s future—and its ability to keep pace with rivals on the document side, most notably MongoDB.

The Couchbase CEO says that the key to the Elasticsearch integration is that it appeals to some of their bread and butter customers in the media industry who want to mesh Couchbase with full text search capability. Further, many of these media folks want to store their metadata in a database like Couchbase for faster access but they still want to be able to do comprehensive full text searches across documents and other data.

The company claims around 350 customers in everything from Web commerce, media and advertising, to some newer areas that are at the higher end of enterprise, including financial services companies like Experian. Wiederhold says that enterprise is picking up significantly—last year it accounted for just 1/3 of its business but it’s moving closer to double that.

This is all part of a shift that is taking place as traditional enterprises, which got something of a late start with their existing infrastructure, have been forced to look beyond the relational wall as their applications are growing in complexity.

Wiederhold touched on the changing nature of data and the role of NoSQL, noting that 20 years ago a website was just static content, however, now they’re full-fledged complex applications (in many cases) and so they’re increasingly moving that stuff over to NoSQL.  The thing is, a lot of them that developed these kinds of applications in like the late 90’s. At the time, there was no MySQL, there was no Postgres, but of course, there was certainly was no NoSQL. Naturally then, a lot of those things were built on top of Oracle or DB2, but as Wiederhold says, “we’re starting to see those guys move off of that from a cost perspective as well as because NoSQL provides the ability to support much greater number of users, much more data, etc.”

He reminded us that “The database industry is a $30 billion dollar industry – 95% of that, today, is where Oracle is from.  So they’re focused on that.  They see that other 5% as relatively uninteresting.  And, we think this is the future.” The CEO noted that on top of that, applications that are being developed today are cloud based using the internet architecture of a browser, a mobile app connected over the internet to either a public cloud or a private cloud– that’s how applications are being developed today.  So at this point, Couchbase is very happy playing in that 5% today, and watching that 5% grow to 7%, then 10% and so on.

Related Articles

New Ropes for Scaling the SQL Wall

MapR Traces New Routes for HBase

HadoopWorld Special: This Week’s Big Data Top Ten