While some see the mainframe as a relic, others recognize its place as the workhorse of future computing. The mainframe stands to play a pivotal role in the age of Machine Learning.
The Evolution of the Mainframe
“The mainframe is dying.” I can’t remember when the first time was that I heard people saying that, or how many times I have heard something similar since. Suffice it to say, it was a long time ago and I’ve heard it many times, and as we all know, the mainframe is alive and thriving.
In fact, the mainframe has evolved to ensure its foundational role in new trend after new trend. After going through the era of distributed computing, we are now experiencing the era of mobile computing and cloud computing. And through all this, the mainframe is still there, playing its central role as the hub of customer information and high volume transaction processing at many of the world’s largest companies.
That said, today’s mainframes face new issues and challenges, top of which are security and automation of operations.
Does Mainframe Security Go Far Enough?
Why do I think security is an issue for mainframes? More than 80% of corporate data resides on or originates from mainframes now. And data on mainframe is typically the most valuable data for the enterprises.
Mainframes used to be perceived as being highly secure and immune to attacks and data breaches because they are centralized and somewhat isolated from the rest of the world. In fact, in its recent “State of the Mainframe” survey of 187 IT strategic planners, directors/managers, architects and administrators at global enterprises with $1B or more in annual revenues, Syncsort found that 83% of respondents cited security and availability as key strengths of the mainframe respectively. But strangely, this is no longer an accurate perception, times have changed.
In today’s world of connecting “Big Iron” to “Big Data” for advanced business analytics, mainframes are connected to mobile and IoT devices, clouds and other open systems, and therefore subject to external attacks just like distributed systems.
At the same time, mainframes are also subject to internal attacks such as employee’s malicious intents or negligence. Although z/OS mainframes remain the most secure platform compared to distributed systems, due to the high sensitivity of the data stored on them, their security is always a major concern.
The Challenges Driving Automated Mainframe Operations
Why is automation of operations important for mainframes? In order for mainframe machines to run smoothly and efficiently, a lot of operations are needed every day, including trouble-shooting issues and optimization of resources.
Historically, most of these jobs have been done manually by human operators. This worked fine in the past, but as the number of transactions grows exponentially, the demand on machine operations and maintenance equally increases. At the same time, the population of staff with mainframe skills is shrinking as senior mainframe engineers retire at a rate greater than young mainframers are being hired. While efforts from groups like SHARE.org are helping, the skill gap keeps expanding. The only foolproof solution to this problem is the automation of the operations of mainframe. Ultimately, automation could lead to self-management of mainframe systems.
A lot of work has been done to address both of the issues described above. Some progress has been made, but not as fast as the increase in complexity we are seeing.
While new security features have been implemented on the mainframe, there have also been new and even more sophisticated threats. And while some operations have been automated, the activity patterns change so quickly that the program is soon outdated.
Apparently, the traditional methods are not fast enough to keep up with the pace. And in some cases, they are even totally incapable. Is there a way out? What if the machine can learn and adapt by itself?
The high value and sensitivity of mainframe data is driving new automated operations and security challenges for IT organizations.
Enter Machine Learning
Machine learning is familiar concept to many. With the victory of AlphaGo, a computer Go program developed by Google DeepMind, against the top professional human Go player, played in Seoul, South Korea between 9 and 15 March 2016, “machine learning” suddenly became a hot topic for the general public.
Although it has the impression of being trendy and cutting-edge, machine learning is actually not a new concept. The early research of machine learning dates back to as early as 1950s, and it has come a long way since then. For many years, machine learning had been more of theoretical studies than practical applications. Only in recent years, backed by new powerful computing technologies, has machine learning started to get into areas that have an impact on our daily lives.
Today, machine learning helps human beings in many different areas. Self-driving cars, computer-aided diagnosis, and the above-mentioned gaming program are just some examples. If you get an email recommending a product that happens to be exactly what you need, it’s very likely there is machine learning behind it. The question remains, in addition to these cool applications, can machine learning help enterprise executives? Can it help us mainframe engineers?
To learn about what machine learning is and how it can help us, read Zhe (Maggie) Li’s eBook: Mainframe Meets Machine Learning.