I have spent my career working with computers. I still enjoy using new gadgets or learning how something like AI (artificial intelligence) works. Yet, not everything in tech is new, even in computers. I recently learned that mainframes, a relic from the 1950s, are alive and well, and in fact, their sales are increasing, presumably because they are quite suited for AI. Mainframes and AI are not words that should coexist! Here’s what has happened.
I learned to program decades ago writing in Fortran. Fortran was one of the first high-level programming languages, written before I was born. It was designed primarily for scientific and engineering calculations.
My programs ran on an IBM mainframe, a powerful computer that easily processes huge amounts of data, known for their speed, security and reliability. The name comes from the large cabinets (main frames) they were once stored in. Mainframes could weigh tons, living in large, temperature and humidity controlled spaces with maintenance personnel always onsite.
I also knew COBOL, another early high-level programming language, designed for business applications. Since then, I've programmed in an endless number of languages and hardware configurations. Both grow like weeds, each addressing some special situation or need. Since their start, new notable programming languages have been developed at the rate of about one a year.
Over time, mainframes dropped off in use as computing moved to servers, personal computers and smart phones, the hardware now generally used in today's internet.
I’ve known for years that COBOL has some residual holdings in business applications such as payroll processing, most notably Walmart. But I was surprised when I heard of someone who had recently used Fortran.
I continued searching, and yes, what goes around may come around. Mainframes, Fortran and COBOL are all alive and well.
Mainframes were the first commercially available computers. Because of their expense, originally they were only available to large organizations, including the government. They are essentially "black boxes," an environment that does not interconnect well with the outside world, a sharp contrast to today's tech environment that is essentially an endless interconnected network.
But this closed design that encapsulates processes has some benefits. It is easier to control both the hardware and the software, giving it a high degree of reliability, availability and security. Further, because most of the processing happens in a physically small area, processing can be very, very fast, able to complete computations in a fraction of the time they may take running on multiple servers and workstations, a common configuration today.
So, yes, there are not just a few mainframes hanging around but instead they are a viable industry, nearly all run by IBM. A majority of the companies in the S&P 500 have at least one mainframe, and many industries such as banking, airlines, insurance and government rely heavily on mainframes for their core functioning.
And now they are coming into use for AI, primarily because of their core competencies: fast, reliable and secure. And they can be cheaper to run, too.
I also found that millions of lines of COBOL code runs daily. It's estimated that more than half of all business transactions globally are processed using COBOL. And for similar reasons: It's fast, secure and it works.
Many of these business processes still handled by old languages have changed little. Payroll does about the same as it did fifty years ago: hours, rate, total pay, deductions and net pay. Process once every week or two, or once or twice a month. In payroll, there's really nothing new under the sun.
And what about my Original Language, Fortran? It is still widely used, a top choice for heavy number-crunching in scientific and engineering fields such as meteorology and physics. It has survived for some of the same reasons other old technologies are around: It is fast, has lots of libraries accumulated over decades and there’s just not a compelling reason to move to something else.
To be clear, these ancient tools have changed dramatically over the years. A mainframe today, almost exclusively built by IBM, is millions of times faster than their forebearers, and much smaller. Their architecture has changed little but the individual components are made using the most modern technology.
Although less so, old languages still in use today often have many of the modern constructs so common in newer languages, such as object orientation, parallel processing and recursion.
The world of technology changes at lightening speeds. New ideas and possibilities are endless. But like so much in life, there are functions and needs in technology that were solved decades ago, and there hasn’t been a compelling reason to change.
There might be a life lesson buried somewhere here.
I learned to program decades ago writing in Fortran. Fortran was one of the first high-level programming languages, written before I was born. It was designed primarily for scientific and engineering calculations.
My programs ran on an IBM mainframe, a powerful computer that easily processes huge amounts of data, known for their speed, security and reliability. The name comes from the large cabinets (main frames) they were once stored in. Mainframes could weigh tons, living in large, temperature and humidity controlled spaces with maintenance personnel always onsite.
I also knew COBOL, another early high-level programming language, designed for business applications. Since then, I've programmed in an endless number of languages and hardware configurations. Both grow like weeds, each addressing some special situation or need. Since their start, new notable programming languages have been developed at the rate of about one a year.
Over time, mainframes dropped off in use as computing moved to servers, personal computers and smart phones, the hardware now generally used in today's internet.
I’ve known for years that COBOL has some residual holdings in business applications such as payroll processing, most notably Walmart. But I was surprised when I heard of someone who had recently used Fortran.
I continued searching, and yes, what goes around may come around. Mainframes, Fortran and COBOL are all alive and well.
Mainframes were the first commercially available computers. Because of their expense, originally they were only available to large organizations, including the government. They are essentially "black boxes," an environment that does not interconnect well with the outside world, a sharp contrast to today's tech environment that is essentially an endless interconnected network.
But this closed design that encapsulates processes has some benefits. It is easier to control both the hardware and the software, giving it a high degree of reliability, availability and security. Further, because most of the processing happens in a physically small area, processing can be very, very fast, able to complete computations in a fraction of the time they may take running on multiple servers and workstations, a common configuration today.
So, yes, there are not just a few mainframes hanging around but instead they are a viable industry, nearly all run by IBM. A majority of the companies in the S&P 500 have at least one mainframe, and many industries such as banking, airlines, insurance and government rely heavily on mainframes for their core functioning.
And now they are coming into use for AI, primarily because of their core competencies: fast, reliable and secure. And they can be cheaper to run, too.
I also found that millions of lines of COBOL code runs daily. It's estimated that more than half of all business transactions globally are processed using COBOL. And for similar reasons: It's fast, secure and it works.
Many of these business processes still handled by old languages have changed little. Payroll does about the same as it did fifty years ago: hours, rate, total pay, deductions and net pay. Process once every week or two, or once or twice a month. In payroll, there's really nothing new under the sun.
And what about my Original Language, Fortran? It is still widely used, a top choice for heavy number-crunching in scientific and engineering fields such as meteorology and physics. It has survived for some of the same reasons other old technologies are around: It is fast, has lots of libraries accumulated over decades and there’s just not a compelling reason to move to something else.
To be clear, these ancient tools have changed dramatically over the years. A mainframe today, almost exclusively built by IBM, is millions of times faster than their forebearers, and much smaller. Their architecture has changed little but the individual components are made using the most modern technology.
Although less so, old languages still in use today often have many of the modern constructs so common in newer languages, such as object orientation, parallel processing and recursion.
The world of technology changes at lightening speeds. New ideas and possibilities are endless. But like so much in life, there are functions and needs in technology that were solved decades ago, and there hasn’t been a compelling reason to change.
There might be a life lesson buried somewhere here.