BOSTON -- Startup NuoDB Inc. on Tuesday released its NuoDB Cloud Data Management System, which is said to bring the elastic scaling of cloud computing architecture to SQL database technology. As opposed to NoSQL
Cambridge, Mass.-based NuoDB's goal is to bring geographically dispersed distribution along with SQL transactions to databases in a way that takes advantage of massively scalable cloud computing systems. With it, servers can be added incrementally to address data processing needs.
This cloud database engine comes into a world that includes a variety of upstart products based on non-SQL and new SQL schemes. Such methods were made popular by Web powerhouses, such as Google and Amazon.
In the face of Web-scale data intake, Google, Amazon and others needed to try new approaches to data. Their non-SQL applications set the tone for the slew of advanced technologies.
"We are talking about fundamental changes in volume and velocity," said Matthew Aslett, research manager of data management and analytics at London-based 451 Research Group. "The likes of Facebook, Google, Amazon and others could not rely on the exiting relational database vendors, because they were locked into the 35-year-old [relational database] RDB architecture. They invented the NoSQL space."
But, even in the powerhouses' backrooms, some SQL-oriented transactional relational database management system (RDBMS) capabilities are needed. Some of the poster kids of NoSQL give proof. "Google is doing a lot of SQL inside," Aslett said. "They needed ACID [for Atomicity, Consistency, Isolation and Durability]." He coins a term to describe the quest: "transactionality."
Under the hood of a new SQL engine
At its heart, NuoDB is an object-oriented, asynchronous peer-to-peer database based on individual units known as "atoms." In a way, the software combines aspects of modern messaging systems with basic SQL programming needs. Add to that the resiliency and failover that are highly valued on Web server farms and the cloud. Overseeing operations is a NuoDB Transaction Engine, which is a process that executes a SQL layer comprising atoms. The software is said to run on Amazon, Rackspace and other cloud computing architectures.
"We need a scale-out architecture on the Web. Relational databases no longer scale," said Gary Morgenthaler, a partner at Menlo Park, Calif.-based Morgenthaler Ventures, a NuoDB venture capital backer, and a veteran of notable database concerns, including the former Ingres Corp. and Illustra Information Technologies. While NoSQL has potential for some types of applications, SQL approaches have great value in transactional settings, Morgenthaler said. "We don't have to throw out ACID transactions and SQL," he told a crowd gathered here for the rollout.
For more on advanced databases and related
Find out how Hadoop will coexist with relational data in 2013
Read a definition of NoSQL (Not Only SQL)
Learn how an effective 'big data' strategy helped an advertising firm attract clients
NuoDB is led by CEO Barry Morris, who served previously as CEO at Iona Software and Streambase Systems. "We are at a major changing point, going from client-server [computing] to what you could call cloud," Morris said.
Like other cloud-based software, cloud databases need to adhere to some unique requirements, one of which is the ability to "soldier on" despite sometimes flaky connections. "You have to design a cloud data management system for failure," he said.
For potential users, NuoDB's projected ease of management of advanced databases on the cloud is of much interest. Early evaluator Dave Tewksbary, technical expert at Dassault Systemes based in Waltham, Mass., said his firm took NuoDB through a very thorough review, and "we scaled it up just by right-clicking."
The Dassault products that Tewksbary's Enovia division creates include product and project design applications. The tests Tewksbary's group ran included data types ranging from binary large objects to bills of materials, and "thousands of SQL statements" per transaction. This was a full-fledged demonstration, "not a shopping cart," Tewksbary said.
In the Dassault evaluation, Tewksbary tested two- and four-node setups -- far short of what people have come to associate with massively scalable cloud setups. At the NuoDB rollout the company demonstrated 1-million-plus-transaction-per-second performance for a job running on six nodes, and higher performance with as many as 12 nodes applied, again, with simple console-side clicks.
This and other moves betoken significant changes are ahead. In the future, said 451 Group's Aslett,"the data management systems on the back end are not going to look anything like those of today."