Business Cloud News
NewSQL is an emerging category that borrows some of the scalability benefits of NoSQL databases for the relational database world

NewSQL is an emerging category that borrows some of the scalability benefits of NoSQL databases for the relational database world

NuoDB, a Massachusetts-based vendor of NewSQL technology has announced the opening of its UK headquarters this week. Martin Gaffney, NuoDB’s newly appointed EMEA regional director told BCN that the rise of NewSQL and cloud combined with the relative lack of innovation in transactional databases has created a perfect storm for the firm.

The company, which was founded in 2008 as NimbusDB and changed its name to NuoDB in 2011, set up its EMEA HQ in Windsor this week, and appointed Martin Gaffney (formerly of Netezza, acquired by IBM in 2010) to head up the new office as EMEA regional director.

“The market space we operate in is very much one that is largely serviced by the legacy vendors, people like Oracle, Ingres, Sybase (SAP), IBM DB2, those types of companies. They’re 20 to 30 years old as organisation,” Gaffney said.

While the company hasn’t garnered much media attention, NuoDB has notably managed to accumulate some top talent. The company’s board of directors includes the founder of Informix (Roger Sippl), the founder of Ingres (Gary Morgenthaler), and one of the early chief executive officers of Sybase (Mitchell Kertzman).

NewSQL is essentially relational database architecture with some of the scalability properties of NoSQL databases. NuoDB’s vice president of technology Boris Bulanov (formerly of Informatica) told BCN the technology promises geographical-distributable, multi-tenanted, highly available and replicable database deployments with the familiarity of SQL – including ACID-compliance.

“On the one hand we are squarely designed for OLTP and have all the characteristics of relational databases, but traditional approaches to relational databases are monolithic – meaning whether it’s Oracle, DB2 or Sybase they are all designed from the same template: they’re all scale-up architectures,” Bulanov explained. “NuoDB has a scale-out as well as scale-up architecture, where you can add commodity resources to improve relational database SLAs – notably throughput and latency; when you look at traditional SQL databases, they don’t scale out.”

He said the simplest way to approach NuoDB is to think of it as a three-tiered architecture, like many cloud-based apps sitting in virtualised environments, where every tier –application logic tier, database tier and the storage tier – can scale out. Splitting the database and storage tiers, which are typically tightly coupled in a traditional SQL environment, is key to making the two independently scalable (and fault tolerant) as well as immediately consistent across all of its elements – and thus very appropriate for cloud-based high frequency OLTP-type applications.

The company has around a dozen enterprise customers (350 individual customers) and announced a couple of recent customer wins including UAE Exchange, an Abu Dhabi-based financial services firm, and IT services provider 1E, and it intends to target large enterprises in three sectors primarily: financial services, telecoms and manufacturing.

Gaffney said that NuoDB’s proposition is that it’s specifically designed for the cloud, which fortunately coincides with a massive increase in cloud adoption within the last few years. The challenge of course is that NuoDB’s use cases primarily pit it against a few large incumbents that already claim the lion’s share of the OLTP market – not to mention other NewSQL vendors like Clustrix and GenieDB.

“We have something designed for the modern environment, and not something designed 30 years ago,” he said. “There’s a level of concern about existing suppliers, but also about the move to cloud-based architectures.”

  • Simon January 25, 2015 at 1:58 am

    This is largely just marketing by Bulanov. He misrepresents the facts. For example, DB2 for Linux UNIX and Windows has always supported scale out, using shared nothing architecture up to 1000 nodes.

    That said, shared nothing has disadvantages, which is why IBM offers pureScale for OLTP workloads – but either is an option.

  • Post a comment

    Threaded commenting powered by interconnect/it code.