Cutting Big Data Down to Size

Last week, I wrote about a choice that could need to be made between cloud resources and the Hadoop tool for managing and working with big data. A better question, or a better way to frame the debate, could be as a decision about what is the best way to make use of cloud computing for data management, especially for "big data."

Tim Vogel, a veteran of data management projects for several major investment firms and service providers on Wall Street, has focused views on this subject. He advises that the cloud is best used for the most immediate real-time data and analytics. As an example, Vogel says the cloud would be an appropriate resource if one was concerned with the most recent five minutes of data under volume-weighted average pricing (VWAP).

"The cloud isn't cheap," he says. "Its best use is not for data on a security that hasn't traded in two weeks. Unless the objective is to cover the complete global universe, like [agency broker and trading technology provider] ITG does." Vogel points to intra-day pricing and intra-day analytics as tasks that could be enhanced, accelerated or otherwise improved upon through use of cloud computing resources. Data managers should think of securities data in two layers—a descriptive or identification layer and a pricing layer—both of which have to be processed and filtered to generate usable data that goes into cloud resources.

The active universe of securities as a whole, which includes fundamental data and analytics on securities, is really a super-set of what firms are trying to handle in terms of data on a daily basis, observes Vogel. With that in mind, the task for applying cloud computing to big data could actually be making big data smaller, or breaking it down into parts—cutting it down to size. That certainly will cut down on the bandwidth needed to send and retrieve data to and from the cloud, and consistently reconcile local data and cloud-stored data.

If nothing else, this is certainly a different way of looking at handling big data. It is worth considering whether going against the conventional or prevailing wisdom could lead data managers to a better way. Inside Reference Data would like to know what you think about this. We've reactivated our LinkedIn discussion group, where you can keep up with new stories being posted online, live tweets covering conference discussions, and provide feedback to questions and opinion pieces from IRD.