Why MongoDB Is Not Just For Big Data
MongoDB is the go-to NoSQL database for Big Data and small projects alike, says 10gen CEO Max Schireson,
MongoDB is the leading database in the open source NoSQL movement, which rejects the orthodox approach of relational databases. NoSQL databases are not built on tables, and tend not to use structured query language to manipulate data.
The database boasts more than 2.6 million downloads and 500 global customers including O2, Craigslist, Disney, eBay, Forbes, foursquare, Intuit, Shutterfly, Telefonica and UK Government Digital Services. The software is supported and developed by 10gen, whose CEO Max Schireson told TechWeekEurope the secrets of MongoDB’s success, on a visit to the UK to open the firm’s London offices – which are, of course, in “Silicon Roundabout”, near Old Street, London.
As well as technology, MongoDB’s business model gives it a head start, Schireson told us. It emerged as a technological response to the massive needs of serving adverts through the Doubleclick system (now a subsidiary of Google), and an early decision to go open source freed it up for the emerging worlds of cloud and Big Data.
MongoDB got its start after ad serving company Doubleclick was sold to Google in 2006. Doubleclick leaders Kevin Ryan, Eliot Horowitz and Dwight Merriman had seen how vital the database was to Doubleclick,and also noted the need to improve on it. ” DoubleClick technology is still in use,” explained Schireson. But the job of serving ads has scaled massively. Doubleclick’s database wasn’t capable of handling 400,000 transactions per second on 90s hardware, he said.
Building the database Big Data needed?
Merriman’s solution was to build “the database he wished he had when he did DoubleClick”, said Schireson. So MongoDB (so called because it handles humongous data sets) is horizontally scalable, and works more flexibly than a traditional relational database which can be “an obstacle”.
“Everyone realised five or six years ago that you had to give some functionality to be able to scale horizontally,” Schireson told us. “You can’t do the same things as Oracle UTM and hope it will run fast on a cluster of 100 commodity servers. You have to simplify and take some functions out to get to scale-out.
“Dwight and Eliot realised there is not a straight line function trade off between feature functions and performance… You don’t need to strip away so many functions – just a few key blocking features – and you can build something very scalable and very usable.”
The founders innovated around the data model, and made something which was more productive, agile and natural for developers, he explained. “It scales out for big data, and when developers use it, they don’t experience trade-offs like with other big data solutions.”
Technically, MongoDB differs from other NoSQL databases by using a binary format called BSON to express objects, instead of the human-readable JavaScript Object Notation (JSON) used by most other NoSQL databases.
It’s clever technology, but it hasn’t always been easy for Schireson to sell his good ideas. After nine years at Oracle, he worked at MarkLogic – becoming COO of a company with an Oracle-beating database for unstructured information. That was an uphill struggle, he says, because no matter how good the technology, “we had a traditional enterprise sales model”. Just like previous attempts to overturn the relational model in the 1990s, MarkLogic had trouble closing sales for a new technology.
“You show up and the developer gets really excited, but he has to tell the IT manager, ‘give me £300,000 and I can try the product out, and maybe some day get a volume discount.’ In that situation, the IT manager will say ‘go away’,” said Schireson.
Giving developers what they want
By contrast, the open source model means developers can pick up MongoDB and start using it directly. “Developers are looking for alternative solutions like MongoDB that make it easier to code, scale, and operate applications,” he said. “The hances are people in your team may already be using it.”
10gen gets its money from subscriptions, charged to users who are moving from the development phase to actual production use. These are cheaper and easier to sell than licence deals, says Schireson. As with all open source business models, it involves selling support, and the MongoDB setup makes this easier than some other cases.
“There are different types of open source plays,” he told us. “Some are pure commoditisation: for instance MySQL is only cheaper than the alternative. It is good enough and free, but it is not better than Oracle. Our technology is different and people really like it. It’s more like the Linux play: people don’t use Linux over Windows because it is cheaper – they use it because they think it works better. That is a much stronger place from which to build a business.”
Cost is a big selling point for MongoDB too. When software goes commercial, it can be worrying for IT chiefs. Instead of asking for software to build something, the developers show a finished prototype, built in something the boss hasn’t even heard of – and that will need support. The bosses are cross – until the developer introduces them to 10gen with a subscription which might run to around $20,000 a year for a big implementation.
“In quite a few cases we show up, and IT management say ‘my develoeprs have built this thing, help me to run and support it’,” he said. “They breathe a sigh of relief when they find there is a commercial model.” Unlike traditional licence deals, these can be signed in a matter of weeks.
The subscriber edition has a commercial licence with an indemnity, and a little bit of extra functionality, and that much-needed support. “We sell support for MongoDB with 24×7 guaranteed service levels,” he said. “That’s important when enterprises deploy for mission critical use.”
Getting software costs into line
Pricing is crucial, and he thinks that software costs have got way out of line in recent years. “We don’t think the database should cost way more than the other components of the solution stack. Twenty years ago, you might go and by a Digital Equiopment Vax for a million dollars, and spend a few hundred thousand on an Oracle license. A decade ago, you could spend $100,000 on a Sun box, and still pay Oracle a few hundred thousand. Now, you can buy a Dell box for $6000 – and Oracle still wants a few hundred thousand for its licence!
“That model makes a lot of sense if you are Oracle, but it doesn’t make sense for the industry as a whole. Software should cost a few thousand dollars per server, similar to the cost of hardware, and power and cooling. People would be willing to pay that to a database vendor who adds value to them, and price won’t be a blocker to them using that technology.”
When 10gen is true to its word, it means in some cases the company is “leaving money on the table”. For instance, a UK telecoms provider was struggling for years to make a large project work with Oracle, but got it to work swiftly with MongoDB. “They would have been willing to pay us millions, but that’s not our model,” said Schireson. “In the long run we will have a happier customer base, and they will spread the word among their friends.”
Not just for big projects
For Schireson, it’s not been a hard job moving to 10gen. He clearly “gets” open source, and wants to see it ride the move to scale-out architectures in big data centres. “People are very eager to move to scale-out architectures, and they are looking to move to technology like MongoDB because Oracle doesn’t work on that.”
But it’s not just for big projects. Companies who pick it up for big projects will realise it is productive and agile, so they use it for other applications. “In the UK, O2 built its Priority Moments, mobile offer system and it was natural for them to use MongoDB for that,” he said. “But they wound up using it in other applications across O2 and Telefonica.”
One of these is an integrated product catalogue, which brings together information on rate plans, handsets and accessories, which all run on different back end systems, in different data models. That’s not a humongous heap of data, he explained, but MongoDB let the developers speed up: “A project which was taking a tonne of time and money accelerated very quickly, and it had nothing to do with the Internet or giant data sets.”
So MongoDB is going well – but in open source, how do you know who is the leader? “There’s no Nielsen box to measure usage. Measuring the number of downloads would be no good because there’s nothing to stop us downloading it loads of times ourselves.”
Instead, Schireson gauges it by looking at activity on Google and forums, seeing the skills people claim to have on LinkedIn, and looking at jobs boards. On those admittedly subjective measures, he reckons MongoDB is about four or five times as big as the other main NoSQL player, Couchbase.
There are lots of ways for open source projects to succeed, and MongoDB isn’t the only NoSQL contender. However, its approach to open source success seems to b winning it plenty of friends. Only last week, 10gen leapt into the Wall Street Journal’s “next Big Thing” list of the top 50 tech start-ups, debuting at number nine.
But being the “next” Big Thing is a fragile claim – and many of last year’s top startupa sre falling from favour. What Shireson wants to achieve is for 10gen to be a current Big Thing – and from all appearances, he is well on his way to taht.
How well do you know cloud computing? Test yourself with our quiz!