You are here

Revolution R Enterprise: Faster Than SAS

Thomas W. Dinsmore, Revolution Analytics
Derek McCrae Norton, Revolution Analytics

In analytics, speed matters. How much? We asked the director of analytics from a leading U.S. marketing services provider, a Revolution Analytics customer. Her team supports more than 1,000 predictive models currently in production; her clients expect the team to build a predictive model in 30 minutes or less.

Previously, we shared results from a performance test comparing Revolution R Enterprise (RRE) to SAS.2 That test showed how our unique Parallel External Memory Algorithms (PEMA) technology produces vastly better performance for advanced analytics. Some readers noted that SAS and RRE were not tested running on the same hardware, because the test was limited to a single task. It was also pointed out that SAS offers software to enable deployment in clustered computing environments, similar to what was used for RRE.

We listened—and we set up a new test using the same hardware for both products. To help ensure a fair comparison, we hired an experienced SAS consultant to review the SAS programs, enable them for Grid computing and run the test. We used SAS 9.4 and defined a list of commonly used analytics tasks for the test.

The results:

  • ScaleR ran the analysis tasks 42 times faster than SAS.
  • ScaleR outperformed SAS on every task.
  • The ScaleR advantage ranged from 10 to 300 times the performance.
  • The ScaleR advantage increased when we tested on larger data sets.
  • The new SAS HP PROCs, where available, only marginally improved SAS performance.

In this white paper, we’ll report on the approach we used for the test, together with detailed results.