Thursday, 24 October 2019 20:11

SAS opens up with Open Model Manager

SAS executive vice president, COO and CTO Oliver Schabenberger SAS executive vice president, COO and CTO Oliver Schabenberger

Data scientists put a lot of work into creating models for their organisations, but a surprisingly small percentage find their way into production. SAS is looking to change that.

According to an IDC survey, less than half of organisations say their analytical models are sufficiently put into production, and only 14% say their data scientists' work is fully operationalised

SAS Open Model Manager is intended to help organisations operationalise their open source models so they can better put their data to work.

Part of the problem is the cumbersome manual processes and inconsistent collaboration between IT and business users. The time taken to move models from development to deployment can be significantly reduced by improving model development, production and automation.

SAS Open Model Manager also helps monitor and revalidate the performance of models that have moved into production.

"Organisations have a good handle on building and training analytical models, including open source ones, but there is often a gap when it comes to operationalising those models and pushing them into production, and a lot of the work done by data scientists is lost," according to IDC research director for business analytics Chandana Gopal.

"There is a need in the market for a new generation of model management solutions that allow data scientists to develop models in any language of their choice, and to properly catalog and deploy their analytical models. With this capability organisations can harness the value of their analytical assets and improve transparency through continuous monitoring."

SAS Open Model Manager, to be released in November, provides a single environment for registering, deploying and monitoring open source models written in Python or R.

This helps users compare and assess different models, manage champion and challenger models, and access built-in performance reports to help make retrain/retire/redevelop decisions.

Models can be deployed with just a few clicks, even in different operational environments. Significantly, this includes automatic code conversion from interpreted languages to machine code for greatly improved performance.

SAS CEO Jim Goodnight told journalists that there are too many computer science graduates who only know Python. As it is an interpreted language, it isn't good for production systems unless the developers call external libraries such as those provided by Viya. "Our libraries are all massively parallel" so they can take advantage of modern servers.

A two-socket Intel server can provide 56 cores and 112 threads for around US$25,000, but you need software that can take advantage of this enormous processing power, he suggested.

But getting people to use external libraries is a challenge, Goodnight said. So the company is trying to expose more people to SAS. It is currently used in more than 80 master's programs, and SAS products are available free of charge to students. In addition to the 100,000 downloads of the desktop software each year there are also 90,000 SAS on-demand users (Viya is a server-only product).

"It's just a matter of spending a lot more money on education," because many people assume SAS is the same product they used as students as much as 40 years ago, he observed.

Contrary to that perception, "It's not old... we're probably eight or nine years ahead of where open source [software] is."

For instance, SAS has been massively parallel since 2009, and that change meant a risk calculation that previously took 18 hours to process could be completed in less than 15 minutes.

Open Model Manager improves governance by helping users better understand the function and performance of deployed models over time.

It includes a GUI and a CLI to meet the needs and expectations of different groups of users.

This ModelOps concept helps organisations manage and scale models to meet demand and continuously monitor them to spot and fix early signs of degradation.

The analytics lifecycle – prepare data, explore, model, register, deploy, monitor, retrain, and repeat – is "a never ending process" that can be assisted by SAS Open Model Manager and the existing SAS Model Manager according to SAS senior vice president of business development Tom Fisher.

The company can also help its customers with other software and services, including the ModelOps health check service, a ModelOps framework that can rapidly move model through their lifecycle, and a range of education services, he said.

SAS Open Model Manager will be containerised (Docker, Kubernetes) for simplified deployment in private or public clouds. No other SAS technology is required.

SAS's attitude towards open source has changed, executive vice president, COO and CTO Oliver Schabenberger said. Open source and cloud both provide opportunities for the company, for example to provide more granular offerings such as Open Model Manager, which can be an important integration point for models.

SAS has decades of experience with its multi-vendor architecture that allows most code to run anywhere, with all the vendor-specific code isolated in a smaller part, and this puts the software in good stead to run efficiently on the major public cloud platforms, he said. "We're working with those providers."

Goodnight told journalists that he is developing Schabenberger to become his successor. "He's very good," said the 76-year-old software billionaire, and is currently the number one candidate for the role.

Disclosure: The writer attended SAS Analytics Experience 2019 as a guest of the company.


26-27 February 2020 | Hilton Brisbane

Connecting the region’s leading data analytics professionals to drive and inspire your future strategy

Leading the data analytics division has never been easy, but now the challenge is on to remain ahead of the competition and reap the massive rewards as a strategic executive.

Do you want to leverage data governance as an enabler?Are you working at driving AI/ML implementation?

Want to stay abreast of data privacy and AI ethics requirements? Are you working hard to push predictive analytics to the limits?

With so much to keep on top of in such a rapidly changing technology space, collaboration is key to success. You don't need to struggle alone, network and share your struggles as well as your tips for success at CDAO Brisbane.

Discover how your peers have tackled the very same issues you face daily. Network with over 140 of your peers and hear from the leading professionals in your industry. Leverage this community of data and analytics enthusiasts to advance your strategy to the next level.

Download the Agenda to find out more


Stephen Withers

joomla visitors

Stephen Withers is one of Australia¹s most experienced IT journalists, having begun his career in the days of 8-bit 'microcomputers'. He covers the gamut from gadgets to enterprise systems. In previous lives he has been an academic, a systems programmer, an IT support manager, and an online services manager. Stephen holds an honours degree in Management Sciences and a PhD in Industrial and Business Studies.



Recent Comments