Early mainframe systems filled room-sized metal frames that could occupy between 2,000 to 10,000 square feet. These enormous machines required vast amounts of electrical power, air conditioning and reams of input/output (I/O) devices. Today’s mainframes are smaller than early “Big Iron” machines and are about the size of a large refrigerator. The latest models (for example, IBM z16™ single-frame system with a standard 19” rack) are built to easily integrate with other IT infrastructure and systems in a modern data center, whether that means in an on-premises data center at a company’s physical location or in a cloud data center.

Designed in 1937, the Harvard Mark I, or IBM Automatic Sequence Controlled Calculator (link resides outside ibm.com), holds the distinction as the first mainframe computer. The US Navy Bureau of Ships used this machine during the last part of World War II for military purposes to quickly solve math problems.

In 1951, the Eckert-Mauchly Computer Corporation (EMCC) began building the first commercial mainframe, UNIVAC (link resides outside ibm.com). Soon after, in 1953, IBM introduced its first mainframe designed for commercial business use—the IBM Model 701 Electronic Data Processing Machine. The company’s first electronic computer, the 701 was about 25 times to 50 times faster than its predecessors, with rapid advancements in computing power, memory capacity and smaller size.

Other US manufacturers of large-scale commercial computers during the 1950s included Burroughs, Datamatic, GE, RCA and Philco.  

The first modern mainframe, the IBM System/360, hit the market in 1964. Within two years, the System/360 dominated the mainframe computer market as the industry standard. Prior to this machine, software had to be custom-written for each new machine and there were no commercial software companies. The System/360 separated software from hardware and for the first time, software that is written for one machine could run on any other machine in the line. 

While many associate virtualization with cloud computing, commercial virtualization technology started on the mainframe as a way to logically divide system resources to be shared among a large group of users. Before virtualization, mainframe IT professionals used keypunches, batch jobs and a single OS to carry out IT operations. In 1964, IBM launched the CP/CMS. This lightweight single-user operating system contained the first hypervisor that created virtual machines (VMs), which virtualized the underlying hardware—increasing efficiency and reducing costs.

Introduced in 1970, the IBM System/370 marked IBM’s first departure from magnetic iron ferrite core technology to silicon memory chips to store data and instructions since they produced faster operating speeds and required much less space. Six months after System/370 launched, the phrase “Silicon Valley” first appeared in print in an issue of Electronic News.

Other significant manufacturers in the mainframe market during the 70s and 80s include Fujitsu, Hewlett-Packard, Hitachi, Honeywell, RCA, Siemens and Sperry Univac. During this time, the mainframe industry continued to advance with smaller machines, I/O performance improvements, more significant memory and multiple processors, allowing their functionality and capacity to grow.

In the 1990s, as the use of the personal computer and other technologies accelerated, some analysts predicted the end of the mainframe. In 1991, InfoWorld analyst Stewart Alsop famously said, “I predict that the last mainframe will be unplugged on March 15, 1996.”

Yet the mainframe use survives as a core IT infrastructure across industries. In April 2022, IBM unveiled the latest generation of the IBM zSeries—the z16, featuring the IBM Telum™ processor with industry-first, on-chip integrated accelerators to predict and automate with AI at unprecedented speed and scale (and with extremely low latency).

The modern mainframe contains network, crypto, storage and compression cards with their own processors and memory. It also houses system assist processors (SAP) that speed up data transfer between the operating system and the I/O (input/output devices) and processors for running Linux™, Java™ and other workloads. This setup allows the mainframe to deliver peak utilization continuously while handling high throughput volumes. 

The large number of processors in mainframe technology support businesses across industries (for example, government agencies, utility companies, financial institutions, healthcare organizations) that rely on large-scale transaction processing to handle massive data workloads, high-volume financial transactions and more. Today’s mainframe solutions are also designed to support cloud computing, data management, big data and analytics, artificial intelligence (AI) and quantum computing, with extensions and integration layers that integrate with core systems.

A mainframe acts as a server for storing and processing data at high speeds and can carry out millions of instructions simultaneously. In contrast, supercomputers are much faster, capable of executing billions of floating-point operations in one second. Supercomputers can perform massive calculation-intensive work for weather forecasting, climate research, molecular modeling, physical simulations and more.

Instead, the two systems have merged to form a holistic digital transformation strategy. To that end, modernizing mainframe-based applications has become an essential part of today’s enterprise hybrid cloud approach, which combines and unifies on-premises, public cloud, private cloud and edge settings to create a single, flexible IT infrastructure. 

By integrating and extending mainframe capabilities into a hybrid cloud environment, businesses can choose the best environment for their workloads (whether in the cloud or on-premises) to maximize each platform’s innovations, technical advancements, security and resiliency. For instance, an airline company can create an application for customers to manage their travel information, such as cloud-based reservation information. The service can also access data that are held on mainframes, such as changes in flight arrival and departure times. This service doesn’t replace or improve existing mainframe capabilities. Instead, it allows the customer to draw from the best of both worlds: data residing in the cloud and on-premises.

Application modernization—the process of transforming monolithic legacy applications into cloud applications built on microservices architecture—starts with assessing current legacy applications, data and infrastructure and applying the right modernization strategy to achieve the desired result. While it’s possible to lift and shift applications, restructuring the application to take advantage of cloud-native technologies (for example, containers and Kubernetes) can often deliver more business value.

Today, mainframe application modernization works hand in hand with DevOps, the set of processes and practices for automating the work of software development and IT operations teams. In a recent IDC survey  (link resides outside ibm.com), 82% of organizations reported taking steps to prioritize using the same application development tools across mainframe and cloud-native environments. Development teams are leveraging DevOps and DevSecOps practices and delivering applications through automated and integrated pipelines for faster, more agile delivery of software releases and updates across targeted hybrid cloud environments.

Sustainability has become a business imperative. With data centers accounting for approximately 1% of global energy consumption (link resides outside ibm.com), large organizations are looking for ways to reduce IT energy usage as part of their ESG initiatives. Modern mainframes use less energy and have a reduced real estate footprint, helping to improve data center efficiency.