At the SAS Global Forum in April CEO Dr. James Goodnight unveiled a stunning new platform for SAS, called SAS Viya. The demos of Viya were amazing, but could the actual product stand up to rigorous testing by power users? Prescient’s Dr. Andrew Kramer got to do an early test drive of Viya and his report is given below.
I was given access to an Early Preview version of SAS Viya, which enabled me to evaluate this fundamentally different environment in which to develop and deploy SAS programs. Viya represents more than just an integer upgrade. I’ve been using SAS for 36 years and this is the most substantial facelift that SAS has undergone.
(The version of Viya that I previewed is embryonic. There will be many more features and procedures added, even by the time you read this article. But here are my early impressions.)
SAS is now in the cloud. Hallelujah! Machine learning programs run fast under Viya. Really, really fast. I ran a back propagation neural network on a data set with 600k rows in 40 seconds. From home. That’s right, I logged into Viya using a Google Chrome browser on my laptop at home; there was no SAS install on that machine. Yet I was able to execute neural networks and then random forests algorithms while kicking back before dinner.
SAS on steroids
Perhaps the best way to describe Viya is that it’s a cross of SAS Studio and SAS Enterprise Miner, with the resultant offspring being on steroids. The Windows Display Manager’s user interface is gone. Instead, SAS Studio is the new living quarters. So if you’re hesitant about leaving Display Manager, the ultimate move to Viya should be impetus enough to make you migrate over to SAS Studio. Besides, SAS Studio is a more accessible way of running SAS, as it includes pre-populated tasks and processes that can generate actual code for you.
Viya comes prepared for serious analytics. This includes copious data pre-processing tools, data mining algorithms, and the ability to utilize legacy SAS statistical procedures. Viya is more “open” than previous SAS architectures, allowing for better integration with languages such as Python, R, and even .NET variants. I did not get a chance to examine the visual analytics portion of Viya, but the examples I saw in April were striking.
A caveat ……. As with any new paradigm, there is a learning curve involved when starting out. Storing and accessing actual code and data sets means utilizing your organization’s cloud. In the Early Preview that I tested this meant servers based at Cary. As a result, I found the concept of a Cloud Authentication Server a bit difficult to initially grasp. Once Viya is in general availability, the ability to use and store my own data sets will make this modality less obscure. Further, seasoned users will need to become acquainted with new PROCS. While the DATA step syntax is largely unchanged in Viya, the PROCs have been substantially modified. Thus it’s important that SAS provides a concise mapping of legacy PROCs to their counterpart(s) in Viya when it’s made available for general use.
If you are new to SAS, then you’ll gleefully embrace Viya. You can run sophisticated procedures without having to know much SAS code. If you’re a SAS veteran, you’ll appreciate how easy it is to test code and then store what works into a library of snippets. Alternatively, you can use procedures from a menu and consequently reuse the resultant code however you’d like. For example, I ran a random forest routine on a specific data set of interest. I decided to perform the same analysis with other data sets, but changing some parameters. I ran PROC FOREST, took the code that SAS produced and made minor changes to have the procedure repetitively do exactly what I wanted.
Viya is SAS’ way of bringing its environment into this decade and well beyond. Now data analysts won’t be constrained by having to write profuse code, instead spending more time being creative.