When the cloud rose to prominence, as usual pundits promptly declared that it would replace previous technologies. In the IT world, anything new tends to capture the imagination, and instantly becomes the former-tech killer. But as time goes on, people figure out that new is not always the be-all and end-all, and old is not always bad, and if we're lucky, we reach an equilibrium.
With cloud, that balance is known as hybrid.
The SAS Institute is proving the point with its new architecture. Dubbed Viya, it's described by its chief architect, SAS analytics server research and development VP Oliver Schaberberger, as "built for the cloud, but it does not require the cloud."
Schaberberger told attendees at the SAS Global Forum, "Cloud is a concept, it's a way of building software that can be deployed anywhere, so you can take SAS Viya into your data centre, bare metal or virtualized, or you can run it in the cloud."
Viya is not a product, it's a platform upon which other SAS products are constructed. This year, SAS plans to deliver Viya-based versions of SAS Visual Analytics, SAS Visual Statistics, SAS Visual Investigator, and SAS Visual Data Mining and Machine Learning. Other products will follow.
While Visual Analytics and Visual Statistics are existing products being ported to the new platform, Visual Investigator and Visual Data Mining and Machine Learning are new. Visual Investigator enables investigators and analysts to reduce false positives by supporting search, query, and visualization of data, including geospatial, network, and temporal visualizations. Visual Data Mining and Machine Learning is aimed at data scientists who want to use machine learning and data mining techniques on all kinds of data, building a model once for use in many places.
All four of the new products use Viya's massively parallel processing capabilities to get answers quickly, and its drag and drop graphical user interface to let users interact with it easily, and with a minimum of training.
According to senior product manager, cloud and platforms Mike Frost, Viya is built on four pillars. It is unified, with a common user interface, management structure, and complete analytics lifecycle. It is open, which, he said, is probably the most fundamental change in SAS's 40 year history, since it allows users to access it via programming languages such as Python, Java, Lua, or REST APIs rather than relying on SAS's own language. It is simple and powerful, letting users generate tasks through a drag and drop interface. And, it leverages the cloud, with support for multiple clouds (for example, Microsoft Azure or Amazon Web Services), elastic scalability, and a pay-as-you-go model.
SAS has wrapped a community around the new platform, where users can collaborate and share code.
The most interesting thing about Viya is not so much what it does, but how it does it. Data is pulled into memory only once (although, if the dataset is larger than the computer's RAM, Viya can handle that too), so processing isn't slowed by I/O. And, it leverages parallel processing – a problem is split into pieces, each of which is run on a separate processor, with results combined to provide the ultimate answer. It's much faster than the standard sequential processing, since a lot of things are happening at the same time.
The magic of the Viya architecture happens when a user combines it with an existing product. For example, a customer can prepare data in SAS 9, which runs in a traditional IT architecture, then transparently pass it off to a Viya product, say, Visual Analytics, for processing. The results are returned to SAS 9, and the user has no idea that the processing has transcended the data centre, moving to the cloud and back. Furthermore, said Frost, the computation is moved to the data, and SAS will deliver software to any hardware that's Cloud Foundry enabled.
Data can be slurped in from multiple sources, including Hadoop, relational databases and streams, and combined as required. And, it is only read once, then retained in memory for analysis, cutting down on the time wasted in reading the same data over and over for different analyses.
Although we talk about "the cloud" as though it were a single entity, nothing could be farther from the truth. There are many, many clouds, and moving between them is often a non-trivial task. Cloud vendors don't like people shuffling data between them. However, Viya happily functions on multiple clouds, including Amazon's AWS, Microsoft Azure, Google CloudPlatform and Oracle cloud.
All of that power is wrapped in a user interface that lets customer perform their analyses by simply dragging and dropping items into the right places. True, they can write code if they want to, but the platform provides for a more Excel pivot table-like interface for users who don't know how, or don't care, to program.
It's one way to cope with the dearth of expertise available. Although SAS (and other analytics firms) are working hard to encourage and enable data science studies, the only solution for the shortage will come over time, as graduates move into the marketplace. No-one can claim that the Viya platform will create instant data scientists, but it will, at least, make it easier for laypeople to manipulate and analyze data.