New SAS Visual Analytics for Big Data
SAS has unveiled a new in-memory analytic platform designed for scalability and very fast performance on standard blade architecture.
SAS Visual Analytics represents a lower entry point than the company's high performance analytic platform, with data exploration and visualization aimed at self-service business users.
SAS CEO Jim Goodnight says the appeal of the in-memory architecture is a combination of time to value in seconds versus hours for business users who no longer have to wait for IT report turnaround, all on lower cost blade servers.
"In memory means you're using every processor you have to work on every bit of the data you have," Goodnight says. "Each processor works on a small block of data that is then read in parallel all at once,"
It's a virtual environment created on the fly from potentially multiple sources of loaded data, not an in-memory database, but data loaded into memory that, in non-relational form, doesn't require the structure of cubes.
A typical Visual Analytics customer might rack install 16, 32 or 48 blades with 96GB of storage and 24 CPUs running on each blade in two threads. (A full rack of 48 blades holds about 3.6 terabytes of data in memory.) Commands arrive via a user's browser and a server that formulates the question and sends it to software that controls the data and divides the work across all the CPUs.
Goodnight says SAS set out to build an in-memory product "capable of running faster than anybody else's in-memory product" and visualization was a likely solution to provide on top. The analytic engine could also support risk calculations, pricing or marketing optimization projects in daily work calculations involving very large amounts of data.
"It's really a platform play if you think about it," says IDC analyst Henry Morris. "One thing I picked up on was the scalability to add more information attributes. You don't want an arbitrary limit on columns, because you have to explore to improve your accuracy and see which attributes about the customer have the most predictive power."
Visual Analytics includes the SAS LASR Analytic Server that calculates data in memory, an explorer for ad hoc discovery and visualization, a designer for custom reports and dashboards, administration for identity and security and a mobile support tool for downloading and viewing reports.
SAS also partners separately for its high-performance analytic platform with Greenplum and Teradata that load data into appliances with distributed code, and those products offer more algorithms and analytic capabilities than the Visual Analytics platform.
Merv Adrian, an analyst at Gartner Inc., says SAS has been looking for more ways of getting analytics closer to the point of a business decision, and Visual Analytics is an example of that.
"They now offer three platform alternatives for in-database, in-memory and grid computing," says Adrian. "It will take time to make all the offerings available in all possible configurations, but SAS’ R&D investment sends the message that it plans to lead its customers into these new architectures."