Accel ponies up $100m to juice big data biz
Using the elephant to mask the elephant
Hadoop World We have entered the business era of the big data eon. That's just a fancy way of saying that software companies are trying to take Hadoop and the myriad new tools that wrap around it and convert them into money by unseating incumbents in data warehousing and analytics – and, quite frankly, by doing some jobs that no one could handle before.
Now the big money is starting to move in: venture capital behemoth Accel Partners is ponying up $100m to invest in big-data startups that are looking to become the application giants that will make a killing by creating Hadoop-based applications that are fit for normal business consumption.
Ping Li, a partner at Accel since 2004, said in his keynote address at the Hadoop World Conference  in New York on Tuesday that the opportunity for making money was enormous. The volume of data being kept is growing very quickly, and every kind of event log or twitch of a company's applications and web servers is being preserved in the hope that it can be mined for insight into customers and markets.
Moreover, unstructured data types are proliferating and companies don't want to rely in the old batch-oriented ways of current Hadoop setups to chew through this data. As usual, business managers want real-time data, just as they did in the 1970s when online transaction processing replaced batch operations for core applications.
This trend is going to mean investment flowing to companies that can improve the Hadoop infrastructure. And unlike those early OLTP days, companies want applications to ride on top of Hadoop and its various adjuncts to provide them with distilled information, and they don't want to build those apps from scratch.
"These new types of data are breaking legacy platforms," Li said. "And this is a big market, not just something for Web 2.0 companies." And to that end, Accel is putting up $100m for its Big Data Fund to help find the Microsofts and Facebooks of the Hadoop application era – and, of course, hopefully before the competition does.
"If you are an entrepreneur with a great idea, I implore you, take some money from this guy," said Cloudera CEO Mike Olson, who also spoke during the keynote. Accel is one of the early investors in Cloudera, which was first out the door with a commercially supported release of the Apache Hadoop stack. "We've got a chance to do something that only comes along once in a generation."
Yahoo! created Hadoop to mimic the operations of Google's search engine circa 2001. It is written in Java and consists of a MapReduce framework that spreads log files, web click streams, and other unstructured data across a network of servers. Then a MapReduce algorithm looks for connections between all this data, as well as any structured data ripped out of database and pumped into the Hadoop Distributed File System.
The mapping part of the algorithm means telling calculations where they need to go to find the pieces of data they need, and the reducing part refers to bringing together the results across thousands of nodes to pull together a summarized answer.
Yahoo! open sourced Hadoop in 2009, and creator Doug Cutting named it after his kid's stuffed elephant. A few Yahoo! employees left to start Cloudera, which jumped out there first to commercialize Hadoop. Since then, the remaining Yahoo! team has been spun out to create HortonWorks, IBM has commercialized its own BigInsights variant, and a number of players including Parascale, MapR, and Appistry have all come out with alternative files systems to HDFS. Platform Computing has even gone so far as to take its own Java-based financial messaging system  used for risk analysis and high speed trading, and make it compatible with Hadoop APIs as an alternative to Hadoop itself.
There's a lot going on out there in Hadoop World. That is just the tip of the elephant's trunk.
Accel was an early investor in Cloudera, which announced $40m in Series D funding at the conference this morning, which it is hosting. This time around, Ignition Partners lead the investment with Accel Partners, Greylock Partners, Meritech Capital Partners, and In-Q-Tel also kicked in some cash to Cloudera. The company has raised $76m in four rounds of funding in the past three years, and Accel has been in all four rounds.
Li says that big-data platforms such as Hadoop – there are others, but they don't generate the same level of excitement – are going mainstream, and that "native" big data applications and services will quickly emerge once companies get Hadoop installed and learn how to use it.
In the prior computing era, where data and systems were not so big – say from 1980 to 2010, in Li's oversimplification – physical mainframes, midrange, and x86 servers were the infrastructure, and applications ran on relational database management systems.
This time around, the infrastructure is going to be virtualized private and public clouds. Hadoop and its add-ons will comprise the data management layer, and a whole new set of analytics, customer relationship management, collaboration, security, and vertical applications aimed at financial services, healthcare, and others areas are going to ride that elephant.
"Multi-billion dollar software companies are going to be created," said Li – and Accel wants to get in on the action.
But before this can happen, Hadoop has had to do a lot of growing up. Hadoop is "a complex piece of multi-tenant infrastructure and it has to behave in your data center," explained Olson.
"It's not enough to have a platform that sophisticated Java developers can use. Business users need to access this," he said.
And over the next year, Olson expects all kinds of apps to come to market to make it easier for regular people – not just Java wizards – to make use of Hadoop. And, if it is done right, they won't even know they're using Hadoop. ®