LexisNexis has built its business on bringing together billions of different records from many different sources. Its data and tools allow customers to query those data to find the names of everyone who registered a car last week in Miami with a license plate that has an "O" and an "H" in it. Recently the company has been working with Sandia National Laboratories to understand whether the LexisNexis data tools might help researchers manage and understand the flood of data coming from the supercomputers and high resolution scientific instruments that drive discovery today.

LexisNexis specializes in data -- lots of data -- about you, me, and just about every other person in the US that has any kind of digital fingerprint. These data come from thousands of databases about all kinds of transactions and public records that are kept by companies and agencies around the US. But just having the data isn't very useful; LexisNexis has to be able to access it on behalf of their customers to help them make complex decisions about what businesses to start or stop, what 500,000 people to send a packet of coupons too, or which John Smith living in California to get a search warrant for.

That infrastructure that LexisNexis uses to do all of this is called the Data Analytics Supercomputer (DAS), and it has been in development and use by the company for a decade supporting its own data services business. The DAS has both hardware and software components, and if you want to host your own internal DAS, it comes as a complete solution ready to run. John Simmons, the CTO for the LexisNexis Special Services Group, explains that while the hardware is based on standard Intel Xeon processors and motherboards from Supermicro, the configuration is specialized to facilitate the rapid processing of very large data streams. So the company specs, configures and assembles the hardware along with the software into a complete system.

A DAS is comprised of some combination of both data refinery and data delivery nodes. These nodes handle the processing of data queries and presentation of results, and up to 500 of them can be connected together by a non-blocking switch (from Force10 Networks, again a standard commercial part) that allows the nodes participating in an operation to cooperate directly with one another. A DAS larger than 500 nodes can be assembled by linking together multiple 500-node sub-assemblies. The system runs a standard Linux kernel with non-essential services turned off to reduce OS jitter and improve performance.

How does it all perform? The company says that in 2008 one of its DAS systems was 14 percent faster than the then TeraSort champion, Hadoop, on a cluster that used less than half of the hardware. Interestingly, LexisNexis also claims that its approach needed 93 percent less code that the Hadoop solution, and this is a big part of the system's appeal.

LexisNexis achieves such work specification efficiency by using its own Enterprise Control Language (ECL), a declarative language developed specifically by the company to allow non-specialist users to construct queries. LexisNexis productivity studies show that ECL is about five times more efficient than SQL for specifying the same tasks, and Simmons gave me an example of a specific data function that was coded in 590 lines of assembly, 90 lines of C, and just two lines of ECL. When you are building a data query engine that has to be accessible by a community of non-specialists, ease of use matters.

There are other commercial solutions in this area, of course. We've written about Pervasive Software's DataRush framework before, as well as IBM's System S. But none of those have the maturity or credibility at scale as the LexisNexis solution.

Even the unimaginative can conjure scenarios in which this kind of capability might be of interest to law enforcement and intelligence agencies, but LexisNexis has been trying something new with its big data engine: scientific data analysis. This week the company started talking about a year-long partnership it's had with Sandia National Laboratories to use the DAS to understand and manage the kinds of very large scientific datasets that high resolution instruments and supercomputers routinely produce.

Richard Murphy of Sandia explained that they have been evaluating how the DAS could fit into the scientific computing workflow. For example, Sandia scientists are experimenting with the DAS in an intermediate step in the workflow to identify regions of interest or high correlation with the occurrence of related phenomena in different datasets. These regions can then be extracted, say for visual analysis, or used as input to different applications in derived computations.

One of the benefits of the DAS that Sandia sees for its users beyond the capability to rapidly process very large datasets is the relative simplicity of the ECL -- scientists can stay focused on their domains yet still construct relatively complex queries of their data without a lot of extra cognitive overhead.

The DAS also has potential for Sandia in analyzing the output of large ensembles of simulations -- as you might find in climate scenario simulations, for example -- all at once, and trying to find features and relationships across the entire ensemble of what could be hundreds of terabytes of output data. Murphy also talked about an application for the validation of scientific codes where the DAS would serve as the engine for comparing computed solutions with data collected from physical experiments. Early results have been promising, and Sandia is making plans for future efforts to take the work further.

Legacy approaches to data management and exploration have begun to sag and split under the strain of soaring data volumes. Mass storage archive systems are capable of preserving petabytes of data but don't help users find it again. And traditional relational database management systems failed at helping us manage the breadth and complexity of scientific data. The LexisNexis solution, and technologies like it that are being developed to deal with petabyte-scale datasets from first principles, offer a departure from established thinking that may finally give us the tools we need to continue turning all those bits of data we produce and collect into information about the world around us.