What if your z Systems platform alerted you that a given strategy had a 92 percent of success? While it can’t do that right now, with IBM’s new cognitive system on IBM z, it could soon.
Your system today represents the peak of programmatic computing. That’s what everyone working in computers grew up with, going back to Assembler and COBOL. Newer languages and operating systems have arrived since; today your mainframe can respond to Java or Linux, Python and Anaconda. Still, all are based on the programmatic computing model. That approach, however, is about to change.
IBM believes that its future lies in cognitive computing. Cognitive computing has become the company’s latest strategic imperative. Security, which became a strategic imperative sometime during 2016, may be the only one left that can rival cognitive computing—at least for now.
The company’s interest with cognitive revolves around data. IBM increasingly sees itself as a cognitive solutions and cloud platform company. To the company, cognitive computing alone will enable organizations to understand the flood of myriad data pouring in. You can’t manually code or define programmatic rules fast enough to possibly keep up.
What you are trying to keep up with is a flood of data: structured data from online transaction processing systems, unstructured data, various social media, Internet of Things, blockchains, surveillance data, images, traffic data from tolling systems and much more. Even your most observant people will have no chance at noticing anything; not even the simplest patterns in this surging river of data. And don’t bother trying to manually spot the subtle patterns that could reveal new and astonishing insights—there isn’t enough time to manually recognize anything, let alone act on it.
In short, you need cognitive computing. It’s how, as IBM puts it,
to move beyond the constraints of programmatic computing; to move away from reliance on structured, local data; to unlock the world of global, unstructured data; to move from decision tree-driven, deterministic applications to probabilistic systems that co-evolve with their users—in effect, learn as they go. Cognitive computing can also take you past a keyword-based search that only provides a list of locations where an answer might (or might not) be located, to an intuitive and conversational means of discovering a set of confidence-ranked possibilities.
IBM has declared this the era of cognitive business. Only cognitive computing will enable enterprises to understand the flood of myriad data fast enough.
Three Paths to Cognitive
If you are an IBM shop, where does that leave you? IBM offers three ways to get cognitive computing solutions: the IBM cloud, Watson or the z Systems server, notes Donna Dillenberger, IBM Fellow, IBM Enterprise Solutions. IBM z, however, is the only platform IBM supports for cognitive computing on premises. It’s also the only IBM platform that supports cognitive natively—mainly in the form of Hadoop and Spark, both of which are programmatic tools.
Even with Spark and Hadoop, you still need an analytics expert. “You don’t program cognitive computing. You give it training models. You train the cognitive system with what you think is valid based on your expertise,” explains Dillenberger. But if you blank on that, don’t worry; in February, IBM introduced a machine learning process in the cloud, starting with z Systems.
Of course, IBM doesn’t expect mainframe shops to completely toss out all they have in favor of cognitive training. “You can still use your old (programmatic) stuff, but the user interface into it is now much more friendly, being front-ended by a new (cognitive) interface,” Dillenberger continued. If you don’t have pre-existing training models, “just use what the cognitive system thinks is best,” she adds.
Machine learning drives cognitive computing. It describes how cognitive computers find patterns. In addition, machine learning can provide algorithms, meaning the system might feed on data from IoT devices or any other sources in search of appropriate patterns. Since a cognitive system continuously learns from data, it can continually refine patterns based on what it has previously seen in through data.
Advantages z Systems
z Systems brings distinct advantages to the cognitive party. First and foremost, notes Dillenberger, “the data resides on z. You can run analytics where the data is and get up to 3x the performance.” Even if you have to pull data from some other location, you still run faster. Other advantages with z Systems include large amounts of memory, multiple levels of cache and multiple I/O processors that get data without impacting CPU performance.
“ETL (extract, transform and load) consumed huge amounts of MIPS. But when an early client did it all on the z, it completely avoided the costly ETL process,” Dillengberger notes. That client reported 7 million to 8 million dollars in savings per year by completely bypassing the x86 layer and ETL and running Spark natively on the z Systems server.
Since IBM just started offering cognitive capabilities on the mainframe, it will take time for the first production users to identify a problem, train their systems and start generating measurable results. However, this might not take as long as you think. Remember, this isn’t a programmatic process. According to Dillenberger, you just need to take the cognitive component out of the box and install it. Your most valuable and insightful data is likely already sitting on the server; you just have to set the cognitive system loose on it.
Cognitive Do’s and Don’ts
If you're going to implement cognitive computing, keep in mind:
- Mobile users will find it convenient to query the mainframe for cognitive answers, which could drive higher z System volumes and charges. IBM discounts mobile-driven monthly software charges.
- You have to be a good data curator. Garbage in, garbage out still matters.
- Use either natural language or a visualization browser to access cognitive. Remember, don’t think programmatic.
One of the biggest advantages of cognitive computing lies in freeing you from the overhead of programmatic computing, but an even bigger payback comes from uncovering business patterns and risks while discovering which strategies will work best. These benefits alone justify the investment in using the z Systems platform with cognitive.
Alan Radding may be reached at firstname.lastname@example.org.