Content Copyright © 2016 Bloor. All Rights Reserved.
Also posted on: The IM Blog
SAS has and is doing a number of interesting things. Perhaps the most surprising, after so many years of focusing solely on SAS code, is that it is opening up the SAS environment to other approaches. For example, its analytics platform includes model management for models built using Open Source R. More recently its Cloud Analytic Services (CAS) – about which, more below – not only includes a RESTful API but explicitly supports native programming access in Python, LUA and Java. This is quite a departure.
As far as CAS is concerned this is a massively parallel in-memory engine that brings together the company’s previous High-Performance Analytics and LASR Server capabilities, and which underpins SAS Viya: the company’s universal cloud-ready architecture for all things related to exploring and understanding data. CAS is interesting in its own right. This is because SAS sees it as an alternative to, and better than, Apache Spark. Given that virtually everybody else is going crazy for Spark this is some position to take. SAS has developed CAS to be especially suited for analytical workloads, which it believes are unique and different from the most common workloads run on Spark, which is a general purpose framework that may be being targeted at analytic workloads but which has not – unlike CAS – been specifically designed to support analytics. Additionally, CAS has advantages when it comes to concurrency, failover and features such as lazy loading from disk. This is a competition that will be interesting to watch going forward.
SAS is doing some other noteworthy developments. For example, it will be introducing a full-blown self-service data preparation product as opposed to the current SAS Data Loader for Hadoop. Initially, the data wrangling offered will be embedded into other products (such as SAS Visual Analytics) – later this year – but it will also be introduced as a stand-alone offering in due course.
More generally, SAS has an interesting take on who uses analytics and for what purpose. While most vendors focus on business analysts on the one hand and data scientists on the other, SAS has identified two other classes of analytic worker, which it calls intelligence analysts and IT analysts respectively. It also recognises two classes of data scientist, one being an analyst per se and the other being a “builder” who is responsible for the process of embedding analytics and algorithms into applications and processes. SAS sees the role of the intelligence analyst as being concerned with how to make analytics actionable while the IT analyst wants to maximise performance and ensure that issues – with performance but also in other respects – don’t recur. The company has released products to support of all of these groups of users: SAS Visual Investigator for intelligence analysts and SAS Environment Manager for monitoring (based on SAS Event Stream Processing streaming analytics) by IT analysts – as well as programmatic and UI based interfaces to SAS analytics for other classes of users.
Finally, it is worth commenting that SAS has a cognitive computing project, which is currently at the prototype stage. Its attitude towards cognitive computing is that it essentially involved deep learning – neural and other networks going into deeper levels than is traditionally the case – together with natural language processing and generation. It is not clear when this will be released.
As I said, SAS is doing some interesting stuff. Moreover, it’s not just developing new products and making itself more open, it is also thinking about its audience in a way that I don’t hear from other vendors.