Too many big data initiatives are science projects that take months of effort, risk failure and require highly trained data scientists with scarce skills. According to a CSC survey, 55 percent of big data projects aren’t completed and many others fall short of their objectives.Read more...
Basho to Bolster Riak with DB Plug-Ins
The simplicity and scalability of a key-value storage architecture has served Basho Technologies well so far. The company behind the open source Riak NoSQL database counts close to 200 customers, including blue chips like Best Buy and AT&T. But moving forward, the company’s new leaders are looking to aggressively augment that key-value base with more sophisticated data storage engines, such as document stores and graph databases, as they take the company to the next level.
Less than two months ago, Basho announced the hiring of Adam Wray, a 20-year industry veteran who spent time at Akamai, to be its new CEO. Wray immediately brought Dave McCrory, another tech vet whom Wray had worked with previously at Warner, to be the new CTO. Last week, while speeding along a Colorado highway enroute to another client site as part of a customer engagement tour, Wray and McCrory spoke with Datanami about the future of Riak, the adoption of NoSQL databases in the enterprise, and what Basho wants to be when it grows up.
“We were brought in specifically because–while the company has an illustrious history from engineering point of view–we have not executed successfully, or really thought through, from an operational point of view, how to answer simple questions, like what does Basho want to be when it grows up?” Basho’s CEO Wray says. “I was engaged by one of the original technical co-founders, and eventually brought on board, and brought Dave and several others with me, to help take the company to the next level.”
Riak is a well-known entity among technologists as a dependable key-value store. The database, which is based largely on Amazon’s Dynamo paper, was written in Erlang, and includes libraries for most popular languages. A technology preview of Riak version 2 introduced last October brought strong consistency enhancements, as well as integration with the Solr search engine.
Riding atop a key value store gives Riak a stronger foundation than other NoSQL databases, claims CTO McCrory. “If you compare us to a lot of the other NoSQL distributed data storage solutions that are out there, we are very strong in the scale out, high-reliability, and easy-to-maintain [areas]. Those are our core strengths,” he says. “What makes Riak such a great story is the fact that it’s based off of a key value store.”
A key-value store is just about the simplest way to store and organize data, besides a barebones file system. That “keep it simple” approach resonates with developers, McCrory says. “For developers today who are trying to write for distributed systems, the more complex the construct that you’re writing to, the harder it is to write to that construct,” he says. “So when trying to deal with some applications, sometimes just a simple key value store is the right and easiest answer.”
The company is currently in the planning stages of rolling out the next phase of Riak’s life, which will transition the NoSQL database into a “data platform.” That plan involves adding new engines to sit atop the key-value store.
“Our goal is to provide you with 80 percent of the constructs you need,” McCrory says. “It could be document store on top. It could be graph. It could be time-series. It could be queuing. It could be caching. If you can think of a construct that’s used to day by the modern Web scaling cloud providers, that’s [what we're looking at]. We’re just not 100 percent sure yet what we’re going to do next.”
That’s why McCrory and Wray are traveling around the country and meeting with clients–to figure out where to take Riak next. Later this summer, the company is expected to announce which additional engines it will add to the key-value store base. It’s possible the company will add multiple engines, but it wants to rank them in order of priority.
Retaining the elegant simplicity of a key value store while building more sophisticated data management constructs on top of it will be a development challenge that Basho’s engineers will dig into heartily. The rewards for building such an extensible data platform are large.
The challenge for Basho is to extend Riak to handle more sophisticated application challenges, without getting too far away from its key-value store roots. “There is no one magic solution today that you can select as an enterprise that will meet all your needs. That’s true of anyone’s product,” McCrory says. “If you try and use a single one, you end up having to do all sorts of unnatural acts to make it effective.”
Keeping the Riak data platform grounded in the key-value store world will give Basho an advantage over competitors, the CTO says. “Many of our competitors have chosen other things as their core constructs, and I think that’s going to expose weaknesses over time,” he says “You can’t make everything look like a document or a column and solve the majority of problems. You can solve classes of problems, but not all of them. And I think that’s going to end up being a problem in the long run for our competitors.”
As Wray sees it, there is plenty of time for NoSQL database vendors to differentiate themselves and grab market share. “I feel we’re not even in the first inning in regards to what the real market potential is,” Wray says. “So somebody who comes out strong out of the gates with a large valuation and hype and noise–the reality is, the difference between my revenues and MongoDB‘s is very little. They just have more hype and larger valuation.”
When it comes to customer engagements, Basho is more often shortlisted beside DataStax, the company behind the Cassandra column-store. When it comes to scaling into the petabyte realm, Wray claims that Riak has an advantage. “You see document stores break down pretty aggressively when you start do to terabytes or hundreds of terabytes of data, let alone much larger than that,” he says. “The operational overhead to shard and handle the support is massive. That’s where we win out.”
These are early, heady days in the NoSQL database market, to be sure. There is a lot of hooting and hollering, and the noise level only seems to increase as more enterprise dollars start flowing into the space. Basho certainly has its hands full as it seeks to go up against larger, well-known, and better capitalized vendors. But the five-year-old company is no newbie when it comes to developing software. If Wray and McCrory can succeed in building new engines for Riak while simultaneously changing the perception of the product as something that only hard-core technologists use, then Basho could become a bigger force in the market.