

A small Cambridge startup founded by two former Akamai executives, and with a new investment in hand, is drawing attention to a database technology space that some web developers and investors believe is the key to letting small startups use cloud-computing infrastructure to take on Internet goliaths.
Basho Technologies Inc. believes 2010 will be the year NoSQL — a catch-all term for developers who reject the popular open-source database software MySQL — becomes widespread on the web. The Cambridge software company’s distributed key values store is winning customers in real-time search and massively multiplayer online casual games and is seeing revenues outgrow projections.
The emerging success of NoSQL is also setting up a new round of fragmentation of the database world, in which customers with heavy online transaction processing requirements stick with technologies such as MySQL, while some media, search and game providers take the leap to the NoSQL architecture, seek systems that stay blindingly fast at a global scale based on less structured data.
“Our planned revenue for (2010) looked like $1.6 million,” said Antony Falco, chief operating officer and co-founder at the startup, which took its first revenue from a new non-relational database product in March of last year. “We should have within the first quarter of (2010) made our number for the year.”
On Tuesday, Basho reported closing a $1.4 million debt financing round, bringing its total investment to $4.7 million. The company is backed by private angel investors and its founders.
In January, Basho announced the appointment of distributed database guru Eric Brewer to its board of directors. Brewer, a University of California-Berkeley professor and founder of Inktomi Corp., wrote the widely credited CAP theorem, which states databases may deliver any two out of three desirable attributes — consistency, availability and partitioning — but not all three.
Brewer said Basho is working on database technology that will enable startups to build web applications that scale like Google Inc. (Nasdaq: GOOG) and Amazon.com Inc. (Nasdaq: AMZN), by relaxing consistency in favor of availability and partitioning.
“Google, Yahoo, Amazon, Facebook, Twitter — all have projects that do data partitioning of some kind on top of clusters,” Brewer said. “But it’s not an easy-to-do thing and it’s certainly not something you can easily use so far if you’re a small group,” he said.
Basho was founded in 2007 by Falco and CEO Earl Galleher, both former Akamai executives who signed on with a sales consulting business to develop a web-based sales tool. In the process, they began building a database architecture using an obscure programming language called Erlang. As the consulting business declined in 2008 and 2009, the company decided to focus exclusively on developing and marketing their proprietary database architecture.
One of Basho’s latest customers is taking on Google in searching the real-time web, an area Google CEO Eric Schmidt has called one of the company’s biggest technological challenges. San Francisco-based Collecta provides a real-time search engine that grabbed headlines in December by offering a live window into goings-on on Myspace.com. The company is using Riak to index and serve up real-time results.
Another customer, Electronic Arts, uses Basho infrastructure for a system that supports 7 million daily users of Warhammer Online on Facebook, saving each player’s status every half-minute, Falco said.
“The indexing systems that are currently out there, they’re not made for these kinds of problem sets,” said Collecta CTO Jack Moffitt. “The problem is that these other solutions are running on the premise that you’ve got this one giant computer in one place with a huge hard disk.”
Web developers and investors say non-relational data architectures are the key that is unlocking cloud computing’s potential for small web developers. With cloud-based infrastructure like Amazon’s EC2, low-budget projects can scale quickly to meet viral ticks in demand, without up-front capital investment in servers.
Traditional relational databases aren’t set up for that kind of need, said Chip Hazard, a general partner at Boston-based Flybridge Capital Partners, who has invested in 10Gen, a New York-based startup with its own non-relational database platform, MongoDB. Applications that involve real-time capabilities and user-generated content run “10 to the fifth better” on a non-relational database, he said.
NoSQL innovators like Hazard and Falco are in competition with one of the database world’s most highly regarded innovators. Michael Stonebraker, known as the ‘father of the relational database,’ for his work on the Postgres project at University of California Berkeley in the 1980’s, points to NoSQL’s weakness: it does not support online transactions. He is working on a Massachusetts-based database project called VoltDB. The project is based in SQL, and Stonebraker says it can equal NoSQL databases in scalability and availability, without sacrificing the consistency needed to run online transactions.
“If you need to run 5 million transactions a second, then nothing around does that,” Stonebraker said. VoltDB will achieve it by relaxing protections against failures related to network partitioning. “We keep (consistency) and (availability) and give up (partitioning),” he said. “Network partitions are unbelieveably rare. The people I talk to aren’t that worried about failure case. When you fudge consistency, [there’s a] huge problem.”
Falco said the majority of web operators disagree. Properties like Facebook and Twitter transact a great deal of information but don’t run very many transactions. “[MySQL databases] are using transactional consistency requirement for every single put and get on the Internet. It’s like shredding paper with a flamethrower,” he said.
Meanwhile, partitioning is extremely important, he said. “What we learned at Akamai is that what people consider as the exception in a network of systems is actually the rule,” he said. “You expect system failures and network partitions. You expect two clusters to be unable to talk to one another for seconds at a time, which is a lifetime.”
The database market is segmenting, Hazard said. Companies with large-scale, transaction-oriented operations, like banks, are likely to continue to use traditional providers, while the new generation of web application developers turns more to non-relational data structures.
“It’s still probably in the early-adopter category,” he said. “All the people we hang around with doing startups are looking at non-relational. We’re starting to see some demand from large, traditional enterprises.”
Editor's note: An earlier version of this story misidentified the database computer language used by VoltDB. VoltDB uses SQL.
Comments
If you are commenting using a Facebook account, your profile information may be displayed with your comment depending on your privacy settings. By leaving the 'Post to Facebook' box selected, your comment will be published to your Facebook profile in addition to the space below.

Print
Email
Print Edition Stories



