First things first: Happy 60th birthday to the IBM mainframe! Did you know that the mainframe was integral to the Moonshot project, initiated in 1961 by President John F. Kennedy? – The platform was initially designed for that very purpose. IBM invested a staggering $5 billion in the project, a massive sum at the time.
Over these 60 years (since 1964), the platform has continuously evolved, known as S/370, S/390, and now IBM Z. It has adapted to industry trends and market demands, supporting modern codes like Java, C, GO, and Python, alongside traditional Cobol and PL/1. Today, it powers technologies for Hybrid Cloud, AI, Analytics, DevOps, Containers, Linux, and Open Source, while maintaining its core principles of Data Integrity, Reliability, Availability, and Serviceability. As the longest-lived technology in the industry, the IBM mainframe’s durability is a testament to its continuous modernization and significant impact on the global economy.
Contrary to the outdated perception of mainframes as “legacy technology,” these systems are anything but obsolete. The mainframe, first introduced in 1964, has continually evolved. Today’s models boast state-of-the-art 7nm chip technology, with a roadmap targeting 5nm chips soon and 2nm by 2030.
Comparing mainframes to the Mercedes Benz S-Class highlights this evolution. Just as the S-Class, first launched in 1972, remains a pinnacle of modern engineering, so too do mainframes represent cutting-edge technology.
Mainframes excel in specific areas, especially in industries like banking. For instance, the majority of the world’s largest banks, including China’s biggest bank and India’s most profitable private bank, HDFC, rely on mainframes for their core systems. HDFC, in fact, adopted mainframe technology just odd ten years ago.
What makes mainframe architecture unique?
In other words, why does mainframe technology remain essential in many industries?
Mainframe architecture excels in vertical scaling, that is: scaling IT production within a highly integrated hardware/software stack. For instance, a fully equipped mainframe with 200 processing units (PU) can achieve 215,089 MIPS (Million Instructions per Second) and handle up to 25 billion encrypted transactions daily!! To put this into perspective, the world’s largest bank processes 1.5 billion CICS transactions daily, with a peak of 14,000 transactions per second[!!!], on a single mainframe with 160,000 MIPS in a high availability environment.
Vertical scaling offers several advantages. Consider a transaction with multiple units of work (UOW), such as a sale requiring revenue booking, receivables entry, warehouse stock adjustment, etcetera across different databases. If any single step fails, it causes an inconsistency. However, on a mainframe, these steps (=CICS transaction) are closely coordinated within a single machine, ensuring both high efficiency and reliability.
Security is another key benefit: up to 25 billion encrypted transactions per day can be processed within a highly secure machine, crucial for compliance and data security.
By the way: While vertical scaling on the mainframe focuses on efficiency and reliability, horizontal scaling typically occurs within a network, mainly in the cloud.
Mainframe Modernization: What’s That? And Do You Get It Right?
First, what does “Mainframe Modernization” mean?
The term “legacy technology” is often indiscriminately applied to the entire mainframe ecosystem. However, hardware and operating systems are not “legacy” but rather cutting edge. The label “legacy” more accurately refers to the application level: these are programs in old programming languages (e.g., COBOL, PL/1), which are often poorly documented and highly dependent on middleware (e.g., CICS, JES). This is precisely where a modernization project should begin: transitioning the legacy code into modern programming languages, including proper documentation. The use of LinuxONE (mainframe architecture with Linux as the operating system) can be advisable in some cases. The typical target language for such modernization projects is often Java – more on that a little further down. For starters, Java runs for free on Z.
In the hype around cloud strategies, “modernization” is sometimes misunderstood and incorrectly implemented: erroneously, the state-of-the-art hardware (i.e., the mainframe) is initially questioned – complex application landscapes designed for vertically scaling hardware are migrated to cloud environments designed for horizontal scaling. Such a migration brings manifold challenges (up to complete failure) because these application landscapes (and the underlying business processes) are not designed for horizontal scaling in the cloud.
What you should know about Java: In the Java Virtual Machine (JVM), the Java program code is always translated to the current hardware. The optimizer or compiler knows very well what works best and always translates the program code into machine language so that the code is executed with maximum efficiency. Generally, Java programs therefore run faster than COBOL programs. This fact is repeatedly surprising for some “mainframers,” as there is an assumption that native code (e.g., COBOL) should run faster. However, a COBOL program compiled 20 years ago does not use new hardware commands; and it is a fact that especially in regulated environments, updates through recompilation on the new hardware do not occur (as often as they should).
Fun Fact: Today, the z-platform is the fastest, most performant Java platform, partly because of the “parallel garbage collection.” The Z Garbage Collector (ZGC) is a scalable garbage collector with low latency. ZGC performs all intensive tasks concurrently without pausing the program’s execution.
I had posted that overview on Mainframe Modernization on LinkedIn, and this triggered quite a passionate debate with many contributions from IT professionals and interesting observations from practice. You can follow that debate on LinkedIN (Click HERE), and I’ll also copy&paste some of those comments / contributions below (w/o name of contributor):
