Revolution links R stats package to apps

Analytics mashup

In the knowledge economy (and boy are we using that term loosely), there are far too many people who need to use analytics than there are people who can create algorithms to crunch data in the open source R programming language. This is one of the reasons why Revolution Analytics, which wants to be the Red Hat for stats, has created a new tool called RevoDeploy R. The idea is to create interfaces that programmers are already familiar with that allow for R routines to be embedded into all sorts of applications.

Revolution Analytics launched back in May, and it's peddling a commercialized and extended version of the R programming language and runtime engine, called Revolution R Enterprise.

In August, Revolution Analytics debuted Revolution R Enterprise V4, which rejiggered the guts of the R engine to allow it to run on multicore and multithreaded processors better and to span multiple server nodes to chew on big data sets using remote procedure calls (RPCs). That V4 release also released a big data format for R applications called XDF, which is loosely based on NoSQL and which allow for data chunking and very high-speed data access to arbitrary rows, columns, and blocks in the store.

The threading and XDF data format are packaged up in a closed-source feature for the R Enterprise V4 software (which is open source) called RevoScale R. This solves a big scalability problem for R, in that data sets can be sucked into XDF and calculations can be spread multiple threads, cores, CPUs, and machines speed up the analysis on big data sets. This is something that the open source R engine cannot do, according to Revolution Analytics.

With the RevoDeploy R feature, which is also closed source, Revolution Analytics has surrounded the R engine with a set of Web services that are implemented as programmable, RESTful APIs. The RevoDeploy R feature also includes Java and JavaScript client libraries that allow programmers familiar with these languages to call R algorithms created by others and without having any particular understanding of R itself.

The feature also has XML and JSON formats for data exchange between the R engine and applications to which analytical capabilities are to be added, and can invoke R scripts anonymously or in an authenticated fashion for higher security. RevoDeploy R includes a repository for storing R objects and RScript execution artifacts so they can be reused by programmers in different applications.

One of the first new users of the RevoDeploy R feature is JasperSoft, which is using the tool to provide instant R statistical analysis and graphing to its JasperServer business intelligence and reporting suite. The tool will also allow, for instance, for quants who work at companies that store lots of data in Excel spreadsheets to embed calls to R algorithms in spreadsheets to chew on data in the sheets.

Revolution Enterprise R V4 runs on Microsoft Windows Server 2003 and 2008 and on Red Hat Enterprise Linux 5; the RevoDeploy R feature only works on RHEL 5 now, and will be ready for Windows soon. Revolution Analytics has been less than forthcoming about the precise pricing for the commercial R tools, but says Enterprise R V4 on a single user working at a workstation costs a few thousand dollars and on a server with a reasonable number of cores and sockets, it's on the order of $25,000 for a license. Pricing has not been announced for the RevoScale R and RevoDeploy R features.

Jeff Erhardt, chief operating officer at Revolution Analytics, says that with the Enterprise V4.1 release, the big data and Web services functionality will be integrated. The company is also looking at add a much-needed graphical user interface to the to the R engine, which should be delivered around the middle of the year or so. ®