DataStax, whose business is based on a distribution of open source NoSQL columnar database Cassandra, has opened its EMEA office in London.
Matt Pfeil, co-founder and vice-president of DataStax, sees Oracle’s recent revenue miss as reflective of a disruption of the traditional, relational database market.
“It’s being disrupted by the cloud and by newcomers – like ourselves – who do things fundamentally differently to the Oracle database in ways that are more efficient for true online applications, where performance and availability matter above all else, especially on commodity hardware”, he says.
On Pfeil’s account, Cassandra, which was developed at Facebook, was “built as an online database first, with 100% uptime. Whether you lose a server or a whole datacentre, the system as a whole is still available.
“Milliseconds matter. Amazon will say that for every 100 milliseconds of latency in the user experience you lose 1% of revenue. Google will say that with every 500 milliseconds of latency you lose 20%. So, downtime is exponentially bad."
Pfeil reflects on Google’s raising of just $25m in 1998.
For more on the NoSQL market
“To index the internet [with so little money], it had to do that on commodity hardware. That approach is now being used across the board," he says.
A database which meets needs
DataStax is the commercial entity behind Cassandra, he says, and offers what it calls: "enterprise-grade security features in DataStax Enterprise (DSE) 3.0". Cassandra is an online database with a distributed, peer-to-peer (P2P) architecture and is used by companies including Netflix, Spotify, Williams-Sonoma, and Pearson Education.
Pfeil says Netflix “decided to move out of their own datacentres and to Amazon Web Services (AWS). They wanted a database that could meet their availability and performance needs on AWS – which is not the most stable of environments."
They chose Cassandra over Oracle, “which likes expensive hardware."
Compared with other NoSQL companies, Pfeil says: “Cassandra’s strength is availability and performance. Being P2P architecturally differentiates it from other NoSQL players. They have a master/slave architecture, which means downtime”.
John Glendenning, who leads the company’s EMEA business, says: “we are about mission critical systems, and are being the most disruptive of Oracle and the ETL/data warehousing world because of availability, scale, performance and economy. We’ve gone from ‘I’m not going to get fired for going with Oracle’, to ‘what does that do for my career?’”
NoSQL = new era, says Couchbase CEO
Meanwhile, Couchbase CEO, Bob Wiederhold is also reading the runes, if not the last rites, over the relational model.
“We are moving on from a 40-year era where relational databases dominated the database world," he says.
“The question used to be: ‘which relational database will we use?’ That changed three to four years ago.
“We are now seeing more NoSQL. That trend will continue. Application developers will pick from a number of database technologies – each of which will have made significantly different architectural trade-offs."
He sees Couchbase’s document-oriented database, again openly sourced, as especially good for “web applications deployed on virtualised or cloud infrastructures and mobile applications requiring real-time data synchronisation between devices and the cloud."
He is sceptical that Oracle’s recent MySQL 5.6, which spans to NoSQL, will trump the NoSQL movement. NoSQL is popular, he says, “because it has made a fundamentally different set of architectural choices that many app developers prefer for the kinds of applications they’re developing today. In Oracle's case, adding a feature to a relational database as a response to what people say they like about NoSQL isn’t going to change these fundamental differences."
Each technology will have its inherent strengths and weaknesses and provide a very different set of trade-offs to pick from, according to Wiederhold. “Sometimes a relational technology like MySQL will be a better fit."
He argues that the ‘big data plus big users’ world of mobile and online gaming companies means that “you need to be able to scale the database tier very easily and dynamically. You can have a million new users overnight and with MySQL it is much harder to scale."
Wiederhold contends that the “infrastructure of the applications being developed today and in the future will be based on open source, whether MySQL or NoSQL." It is, he says, more flexible, and the “freemium approach” allows developers to download and experiment. That is very different to the proprietary software model."
UK and European Couchbase customers include Experian, Sky, Adidas, The Hut Group, Betfair, Unibet, Orange, SkyScanner, Nokia’s Navteq division, Gamegos in Turkey and Scoreloop in Germany.
“We are close to ‘early majority’ in crossing the chasm," he says. “It is still is very young, with only three years of sales for NoSQL”.
10Gen’s MongoDB sees NoSQL as engine of growth
Joe Morrissey, VP of EMEA at 10gen, is a former Oracle executive who led the MySQL business in Europe. He now heads the MongoDB-based company in Europe and is bullish about growth prospects.
“We grew a lot in 2012, and are seeing enterprise adoption, on the continent and in the UK. There is lot of interest in financial services, telecoms, media and publishing and in gaming.
“Organisations are realising that one size fits all is over. The relational model is still a powerful tool, apt for a wide variety of use cases, such as double-entry book keeping. But the scalability, flexibility and agility of NoSQL means the vast majority of growth will be there."
MongoDB is a document oriented database. Morrissey contends that “when customers start to look beyond the relational, the initial driver is often scalability. They’ve usually hit a performance ceiling".
Often then, he says, they come to the “document, or basic unit of record, oriented approach. That solves the scalability problem, but you don’t give up the rich query functionality of relational that you do with a key value store [also non-relational]," he says.
“It allows for iterative business, reducing time to market. Businesses need to be more and more agile and so data models need to follow suit."
The schema building of the relational database model is too rigid, he says, it involves too much work. The document-oriented model, by contrast, “does not store data in rows and columns, it is schema-less and you can iterate quickly.
“This is a time of great transformation in the technology industry," he says.