Basho – the quietly impressive success of Riak
JAXenter heads over to Basho’s London office to talk distributed databases, the rise of NoSQL and building bridges.
The buzz surrounding non-relational databases has reached fever pitch, with the term NoSQL (after fierce resistance) gaining ground in the enterprise world. Much of that battle has been fighting against the term itself however, trying to clarify that the “Not Only SQL” approach isn’t as dangerous as it might seem from the outset.
Another big stumbling block for mass adoption is the fact the NoSQL databases are all lumped under that term, when they can be so very different. The media spotlight has fallen on 10gen’s document-oriented database MongoDB and Datastax’s key-value store Cassandra (getting plenty of attention at Apache too) and naturally the enterprise gravitates towards them.
But unbeknownst to some, there’s actually more specialist scenario NoSQL databases right under your nose. Basho’s Riak is one of them.
Founded in 2008 by a group of developers from Akamai, Basho Technologies were originally a CRM company. The Akamai contingent came to the company with a remit to build the distributed database they always wanted. An easy to manage decentralized database with extreme high-availability, low latency and linear scalability at its heart. Taking inspiration from Amazon’s Dynamo paper and written in the distributed language Erlang, the database that would go on to be known as Riak came about just when developer interest into NoSQL ramped up.
Then came the global recession, taking out a swathe of startups. The people behind Riak soon realised that their newly-created key-value store was where they should place their money. From there, Riak gained popularity in open source realms, before enticing enterprise arms. From their headquarters in Cambridge, MA, Basho has been on a near two-year expansion plan, now employing 100 staff and opening offices in San Francisco, Virginia and London.
Basho’s European offices were opened up in May of this year by General Manager, Matt Heitzenroder and Director of Technical Operations Ian Plosker, with the task of building a Riak community as well as drumming up business interest. [Editor’s Note – neither are still working at Basho]
Heitzenroder detailed the rapid pace in which they set up in Shoreditch, the established tech startup hub of UK’s capital.
“In March, we found the office, secured the lease by April and moved here in May. [It was] Really fast, almost without a plan. We set up the branding on the windows, big glowing Basho logos. It felt like a startup again, in fact it still feels like a startup to me, which is nice. It has the value and credibility of a larger startup, but it is also a smaller startup here in this environment.”
From Angry Birds to Mozilla
By merely glancing at Basho’s client list, you can see they’ve carved a way into enterprise circles impressively, with customers wanting something to solve their key shaped problems. From Angry Birds creator Rovio to Joyent Cloud, Best Buy to the Mozilla Foundation, the variety of use cases is impressive. Even viral social apps like Voxer (which experienced a huge pickup rate after rapper Soulja Boy invited users to rap with him) use Riak to deal with scaling.
Heitzenroder is particularly proud of Riak powering the Danish Healthcare Medical Authority’s datacenters, which services eight million people.
He adds: “Basically when they need to get medical or prescription history for a patient, they need that data as fast as possible because if you’re laying on the table, you’re not awake to talk about it. A doctor needs to be able to access your prescription profile. They need to be able to see it before injecting you with things. I think that’s one really good use case and it’s one that makes me feel really good.”
So with all these enterprises queueing up, why is Riak overshadowed by the big two of Cassandra and Mongo? In truth, it seems it is down to honest marketing, not hiding away from situations in which Riak is not suited to.
“It’s not a fit for everybody and that’s one of the things that made Basho unique. We’re very open about saying what Riak’s not good at,” explains Heitzenroder.
One comparison which prospective customers want to make is with MongoDB, despite the two being markedly different in approach.
“I think one of the reasons why [they are compared], and to discredit Riak, for a veteran/consumer developer, it’s really easy to spin up one single node of Mongo and do a benchmark very quickly with a very productive API,” says Heitzenroder, before exclaiming that Mongo “has a wonderful API” that is “so much richer than a key-value store like Riak.”
Consistency vs. availability
The performance comparison is slightly unfair, due to Riak opting for consistency over availability. Heitzenroder adds that “if Riak is being evaluated against Mongo, depending on the savviness of the engineers who are doing the evaluation, it can be a hard battle because it’s a completely different product.” Heitzenroder believes that in all his time at Basho, he’s never seen a customer leave for Mongo, but sees the reverse a lot, with developers reaching their limits with Mongo.
With a well grounded open source project that has been “legitimised” in the enterprise by the community, Basho now enter different waters with the Amazon S3-inspired RiakCS. The newest product brings them into the cloud storage market , which Heitzenroder says brings different customers to them. The Amazon S3 API acts as a proxy with a Riak cluster underneath, harnessing the powers of the product that made their name. Heitzenroder points out the customer need for lower latency when partitioning in S3.
“For us internally, a lot of the inspiration to build something like this was us dogfooding Riak. Let’s build a product using our product and understand what it’s like to use and build with our product. It was a goal of engineers to gain insight into use cases themselves and also to have something we can go out there and sell.”
The London Basho pack say they’ll continue to build bridges from the Shoreditch office, but suspect that a pilgrimage to Sweden (where Erlang was created) could be on the cards. Either way, Basho’s rise up the NoSQL stakes is worth monitoring.
Heitzenroder thinks that the NoSQL space is thriving at the moment, fuelled by the work done by Google and Amazon. “A lot of Cassandra, Voldemort, Riak wouldn’t exist if it wasn’t for those papers,” says the Basho manager.
“I don’t think we would exist without the Dynamo paper. Where you have Eric Brewer talking about CAP in 2000, Amazon actually applying that commercially and writing a really good paper about some innovation they’d done. That’s inspiring to engineers. I’ve been fortunate to join a bunch of developers who made the right choices surrounding tooling.”