Mainframe: platform of choice for machine learning and ops intel

We have smart homes, smart watches and smart gadgets of all kinds so why not smart mainframe management? Here’s how CA is changing that.

These days, everything around us is getting smarter, from Netflix learning our preferences based on what we’ve watched to cars driving themselves. So, why shouldn’t it be the same for the mainframe management?

Mainframes host mission-essential applications that support thousands of applications and devices simultaneously for thousands of users. Consider these statistics:

  • 70 percent of enterprise transactions touch a mainframe
  • 70-80 percent of the world’s corporate data resides on a mainframe.


The z Systems platform is now the most cost efficient and secure platform to process and analyze all of your enterprise data, from mobile applications to data lakes.  This growing use of data analytics and mobile connected to the mainframe and the transactions volume is adding to tremendous operational pressure and hence the mainframe itself needs to be capable to handling it’s own operational data.

Operational Data: Balancing complexity with performance

How can we cope with the complexity without stalling performance?

With machine learning and operational intelligence on the mainframe, we all could be working smarter, not harder.

That’s why at CA, we are embedding machine learning and advanced analytics into our products such as CA Mainframe Operations Intelligence, which was announced at CA World ‘16.   CA is delivering customer-focused innovation by embedding deep mainframe expertise and self-learning into our solutions, to proactively prevent issues before they happen, and also reduce the dependency on mainframe specialists to identify the root cause of problems when they do occur.

With simplicity and flexibility top of mind, the analytics engine was developed as an operations intelligence software appliance using Docker. This makes it easy for customers to deploy state of the art analytics technology as part of their existing CA software investment. The operations intelligence appliance can also be easily ported between Linux on z Systems, x86 and cloud, helping reduce costs for our customers.

AIOps: Why go on a hunt when “Predicted Time To Avoidance” gives you the answer straight away?

CA is making significant investments in the areas of machine learning, advanced analytics and automation to drive towards more intelligent mainframe management, addressing not only Mean Time to Resolution (MTTR) but more importantly, “Predicted Time to Avoidance” (PTTA).

This represents a shift into a category that Gartner calls AIOps.

“AIOps platforms represent the evolving and expanded use of technologies previously categorized as IT operations analytics (ITOA). This shift is in response to the growing importance (due to digital business demands) and the use of big data and machine-learning technologies across all major ITOM functions, including the service desk, automation and monitoring.” (Gartner, Applying AIOps Platforms to Broader Datasets Will Create Unique Business Insights Published: 01 July 2016 ID: G00296361 Analyst(s): Colin Fletcher)

At CA, we believe that MTTR is just part of the solution because it only alerts the mainframe system operator of an issue after it has happened – reactive problem solving.  With that approach, the challenge is that you have a network operations center with consoles continually showing “as sea of red” and alerts go unnoticed.

And often by the time you see red, it’s too late – the problem has already happened. That could mean, for example, that a customer can’t refresh their bank balance on your mobile app, which in turn could cause you to lose that customer. If you wait for red, you’ve lost the customer experience game – and that doesn’t make your app (and your company) look so smart.

PTTA, on the other hand, provides forecasting to enable ops to act before a problem happens by looking for early signals and then predicting the time it takes to take action to avoid or avert the problem all together.

Whereas other solutions separate the analytics function from the solution, CA embeds analytics into the system view, so mainframe operators can predict a problem before it happens.

For example, if you know that 20 percent of customers’ problems happen 80 percent of the time, the analytics in CA Mainframe Operations Intelligence uses pattern recognition and behavioral algorithms to prescribe the necessary remediation step and automate the entire remediation process.

Thresholds are then automatically generated and applied to ongoing operations based on the normal performance patterns. Alerts are generated when abnormal conditions (anomalies) that exceed thresholds are detected, subject to “smoothing” algorithms to prevent excessive alerting. CA Mainframe Operations Intelligence supports collaborative root cause analysis tools to assist in problem resolution.

Think of PTTA like starting to drink water, taking extra vitamin C and getting extra sleep before you feel that a cold or flu is coming on whereas MTTR is the time it takes to get better after getting a flu and taking medication. The question is, “How quickly can I get back on my feet after a flu, or, in the case of a mainframe, after a slowdown, outage or other performance problems?”

And that’s the difference between smart mainframe operations intelligence and those that aren’t so smart…

Ashok Reddy
Ashok is General Manager of Mainframe at CA Technologies where he's responsible for the P&L,…


Modern Software Factory Hub

Your source for the tips, tools and insights to power your digital transformation.
Read more >
How Regulations Will Impact AI InnovationHow an Internal Incubator Will Optimize Your Dev PracticesWhat You Need to Know About the Serverless Future