IBM is bringing machine learning to a mainframe near you

IBM z Series mainframe

Modern machine learning — often just referred to as “AI” — has largely been the province of specially trained computer scientists who traffic in exotic modeling frameworks and graduate level math. Well-funded startups and large companies alike spend large sums of money to create models that analyze their data — usually with the large-scale resources available in the cloud.

IBM is hoping to change that around in two important ways with its newly announced Machine Learning effort. First, by providing a simplified experience for data modeling and model deployment, it aims to bring those tasks within reach of domain experts already employed by enterprise customers. Second, it has begun supporting deployment of machine learning tools on its z Series mainframes, so they can be run on premises, instead of requiring that data be re-hosted in a cloud like IBM’s Bluemix.

An interface to help data scientists build machine learning models

IBM Apache Spark logoAs powerful as machine learning frameworks like Google’s TensorFlow are, they still require a lot of specialized knowledge to create models. The same is true for run-time and deployment options like Apache Spark. Notebook interfaces like Jupyter provide a way to organize code, but don’t make it any easier to write it. IBM’s Machine Learning system provides a wizard-like interface that lets a data scientists prepare their data for modeling, choose an appropriate model type, evaluate various modeling options, and quickly deploy their models to their IBM servers running z/OS.

Don’t let the name Watson fool you

IBM has cleverly tacked the popular Watson brand name onto a wide variety of its products and services. Many of them have no cognitive computing component at all. In the case of IBM Machine Learning, the situation appears to be a little muddier. IBM says that its Machine Learning offering “extracted the core machine learning technology from IBM Watson.” In plain language, I think that means that the basic building blocks of Watson, like Apache Spark and various machine learning frameworks are now part of its commercial products, but the current version doesn’t appear to include the higher-level cognitive reasoning that was on display during those famous Jeopardy! episodes.

Originally called IBM Predictive Analytics, and based on SPSS, IBM launched Machine Learning late last year on its BlueMix cloud platform. The company has put a lot of work into extending it to newer technologies and tools. In particular, IBM cites that it can now support analytic models in:

  • Any language (including Scala, Java, Python)
  • Any popular machine learning framework like (including Apache SparkML, TensorFlow, and H2O)
  • Using any transactional data type

Today’s announcement means that rather than requiring enterprises to move their data to IBM’s BlueMix cloud and lease computing services, they’ll be able to construct and deploy these applications on their own systems — effectively in a private cloud. This removes the cost, latency, and risk of moving data off premise.

IBM is particularly enthusiastic about the importance of private clouds to industries which manage a lot of sensitive data, like healthcare. For example, Argus Health was one of the keynote speakers at this week’s launch event. They run over 1 billion transactions per year on their IBM z Series mainframes, and are using IBM’s modeling tools to analyze patient compliance and health outcomes to improve the standard of care.

IBM’s z Series will be the first of its computers to get the Watson Machine Learning tools. The company says it will be moving it to other product lines, including its POWER platform, in the future.

About Skype

Check Also

, Motorola Slap Bracelet Phone Prototype, #Bizwhiznetwork.com Innovation ΛI

Motorola Slap Bracelet Phone Prototype

Motorola seems to have called in a bit of the 90s kid spirit with its …

Leave a Reply

Your email address will not be published. Required fields are marked *

Bizwhiznetwork Consultation