ROOT@Z16 > ./mainframe --history --all

IBM Mainframe

The Machines That Ran the World

Before the cloud, before the PC, before the internet — mainframes were computing. Introduced in 1952, IBM's mainframe line has never stopped running. Today they silently power the global economy: banks, airlines, governments, and the ATM down the street.

30 Billion

Daily transactions processed by mainframes worldwide

95%

Of all ATM transactions touch a mainframe

$3 Trillion

In daily financial transactions processed on z-Series

800 Billion

Lines of COBOL still in active production use

// SECTION 02

Timeline of Innovation

Seven decades of computing history — click any card to expand

1952

IBM 701 Defense Calculator

IBM's first mass-produced scientific computer, officially called the Defense Calculator. Only 19 units were built, each leased to government and scientific customers for $15,000 per month. It used vacuum tubes — 4,000 of them — and could perform 17,000 additions per second. The 701 established IBM's commitment to large-scale computing and started the mainframe lineage that continues today.

VACUUM TUBES 17,000 OPS/SEC 19 UNITS BUILT

1954

IBM 650 First Mass-Production Computer

The IBM 650 was the first computer to be produced in quantity — over 2,000 units were built between 1954 and 1962, making it the most popular computer of the 1950s. It used a magnetic drum for main memory (2,000 words of 10-digit decimal numbers) and was aimed at business data processing rather than pure science. IBM's decision to offer it on a low-cost lease brought computing to hundreds of universities and corporations for the first time.

MAGNETIC DRUM ~2,000 UNITS DECIMAL ARITHMETIC

1959

IBM 1401 Launched the Computer Industry

The IBM 1401 was a game-changer: affordable, transistorized, and designed for business data processing. Over 12,000 units were installed worldwide, making it the first computer to achieve genuine mass adoption in business. It replaced punched-card accounting machines in thousands of offices and trained a generation of programmers. IBM's revenues from the 1401 family funded the development of the legendary System/360.

TRANSISTORS 12,000+ UNITS BUSINESS DP

1964

IBM System/360 "The Most Important Product Announcement in IBM History"

Announced April 7, 1964, the System/360 was IBM's $5 billion bet — the largest privately funded commercial project in history at the time. Its key innovation was architectural compatibility: a single instruction set spanning models from a $133,000 desk-side unit to a $5.5 million behemoth, all running the same software. This concept of a compatible family of computers is the foundation of all modern computing. The S/360 introduced 8-bit bytes, hexadecimal floating point, and the idea of an operating system as a distinct product (OS/360).

COMPATIBLE FAMILY 8-BIT BYTE $5B DEVELOPMENT OS/360 IBM System/360 Model 30

1970

IBM System/370 Virtual Memory & Paging

The System/370 extended the S/360 architecture with virtual memory — the ability to use disk storage as an extension of RAM through hardware paging. This allowed programs to be larger than physical memory for the first time in IBM's line. The S/370 also introduced the VM/370 operating system, which could run multiple virtual machines simultaneously — a concept that would not reach mainstream computing until VMware appeared three decades later.

VIRTUAL MEMORY HARDWARE PAGING VM/370

1990

IBM ES/9000 First Water-Cooled Enterprise System

The ES/9000 (Enterprise System/9000) was IBM's answer to the explosion in enterprise computing demand. The high-end models required water cooling — massive refrigeration systems integrated into the machine cabinet — to handle thermal output from thousands of ECL (Emitter-Coupled Logic) chips running at billions of cycles per second. It was the last IBM mainframe family to use bipolar technology before the shift to CMOS.

WATER COOLED ECL BIPOLAR ENTERPRISE CLASS

1994

IBM S/390 Parallel Enterprise Server CMOS Transition & Parallel Sysplex

The S/390 generation marked IBM's transition from exotic bipolar chips to standard CMOS technology, dramatically reducing cost, power, and size while actually improving performance. The Parallel Enterprise Server introduced Parallel Sysplex — the ability to link up to 32 mainframes into a single logical system with shared data and workload balancing. This clustering technology enabled near-continuous availability and influenced modern distributed systems architecture.

CMOS PARALLEL SYSPLEX 32-WAY CLUSTERING

2000

IBM z900 First zSeries — Born 64-bit

The z900 launched the zSeries (z for "zero downtime"), IBM's modern mainframe brand. It was the first IBM mainframe with a fully 64-bit architecture, allowing virtual address spaces of 16 exabytes. The z900 could run thousands of Linux virtual servers alongside traditional z/OS workloads simultaneously, repositioning the mainframe as a consolidation platform for the web era. At 3,000+ MIPS, it outperformed entire server farms at a fraction of the total cost of ownership.

64-BIT ZERO DOWNTIME 3,000+ MIPS LINUX SUPPORT

2015

IBM z13 8 Billion Encrypted Transactions/Day

The z13 was optimized for the mobile economy: it could handle 2.5 billion mobile interactions per day and encrypt all data on the fly without performance penalty — a critical feature as regulations like PCI-DSS required encryption everywhere. The z13 featured a 22-core processor running at 5 GHz, capable of executing over 2.5 trillion database operations per day. It was the first mainframe designed from the ground up for real-time analytics and mobile workloads.

5 GHz, 22 CORES PERVASIVE ENCRYPTION REAL-TIME ANALYTICS IBM z13 Mainframe

2022

IBM z16 On-Chip AI Accelerator & Quantum-Safe Crypto

The z16 is the most advanced mainframe ever built. Its Telum processor integrates an on-chip AI accelerator — the first in any processor — delivering 300 billion deep-learning inference operations per day with sub-millisecond latency. This allows fraud detection on every single transaction in real time. The z16 also implements quantum-safe cryptography (lattice-based algorithms) in hardware, protecting data against future quantum computer attacks today.

ON-CHIP AI QUANTUM-SAFE CRYPTO 200,000+ MIPS 300B INFERENCES/DAY

// SECTION 03

JCL Explorer

Job Control Language — the language that tells the mainframe what to run and where to find the data

What is JCL? Job Control Language is the scripting language of z/OS. Every batch job — whether it compiles COBOL, sorts a file, or processes a payroll — must be described to the operating system in JCL. A JCL job is a stream of statements, each beginning with //, that tell z/OS: who you are (JOB statement), what program to run (EXEC statement), and where the data lives (DD statements). JCL has been the interface to mainframe batch processing since OS/360 in 1966.

//HELLO   JOB  (ACCT),'HELLO WORLD',CLASS=A,MSGCLASS=X
//STEP1   EXEC PGM=IEFBR14
//SYSPRINT DD   SYSOUT=*
//SYSIN    DD   *
  HELLO WORLD
/*

Statement Breakdown

JOB Identifies the job to JES (Job Entry Subsystem). HELLO is the jobname, (ACCT) is the accounting info, CLASS=A routes to an initiator class, MSGCLASS=X controls where job log goes.

EXEC Names the program to execute. PGM=IEFBR14 is a famous IBM no-op — it literally does nothing (used for DD allocation testing). The step name is STEP1.

DD SYSOUT=* Data Definition for SYSPRINT. SYSOUT=* routes output to the same MSGCLASS as the job — typically the system print spool.

DD * Inline data follows immediately in the JCL stream. The /* delimiter marks the end of the inline data.

//COMPJOB JOB  (PROJ01),'COMPILE COBOL',CLASS=B
//COMPILE EXEC PGM=IGYCRCTL,PARM='OBJECT,XREF'
//SYSLIB  DD   DSN=SYS1.COBLIB,DISP=SHR
//SYSIN   DD   DSN=MY.SOURCE(MYPROG),DISP=SHR
//SYSPRINT DD  SYSOUT=*
//SYSLIN  DD   UNIT=SYSDA,SPACE=(TRK,(5,5))

Statement Breakdown

IGYCRCTL The IBM Enterprise COBOL compiler. PARM='OBJECT,XREF' passes compiler options: generate object code and produce a cross-reference listing.

DSN= Dataset Name — the mainframe equivalent of a file path. SYS1.COBLIB is the COBOL runtime library. MY.SOURCE(MYPROG) is member MYPROG inside a PDS (Partitioned Dataset — like a directory).

DISP=SHR Disposition: open the dataset for shared read access. Other jobs can also open it simultaneously. Contrast with DISP=(NEW,CATLG) which creates and catalogs a new dataset.

SPACE=(TRK,(5,5)) Allocate 5 tracks of disk space, with 5 more available as secondary extent. Mainframes allocate disk space explicitly in tracks, cylinders, or blocks.

//SORTJOB JOB  (BILLING),'SORT CUSTOMER FILE'
//SORT    EXEC PGM=SORT
//SORTIN  DD   DSN=CUSTOMER.MASTER,DISP=SHR
//SORTOUT DD   DSN=CUSTOMER.SORTED,DISP=(NEW,CATLG)
//SYSIN   DD   *
  SORT FIELDS=(10,20,CH,A,30,4,ZD,D)
/*

Statement Breakdown

PGM=SORT IBM's DFSORT utility — one of the most heavily-used programs on any mainframe. It sorts, merges, and copies sequential datasets. DFSORT has been optimized continuously since 1966 and can sort billions of records per hour.

SORTIN/OUT Standard DD names expected by SORT: SORTIN is the input dataset, SORTOUT is the output. DISP=(NEW,CATLG) creates a new dataset and registers it in the system catalog so it can be found by name later.

SORT FIELDS The sort control statement: (10,20,CH,A) means sort on field starting at byte 10, length 20, Character format, Ascending. (30,4,ZD,D) means secondary sort at byte 30, length 4, Zoned Decimal, Descending.

ZD format Zoned Decimal is a mainframe numeric encoding where each digit occupies one byte, with the sign encoded in the high nibble of the last byte. It is still widely used in COBOL programs today.

// SECTION 04

Batch Processing

The original computing model — and still the workhorse of global business

📋

SUBMIT

JCL submitted
to input queue

→

⚙

JES QUEUE

Job Entry Subsystem
spools & schedules

→

🔄

INITIATOR

Selects job
by class & priority

→

▶

EXECUTION

Program runs
step by step

→

📁

OUTPUT SPOOL

SYSOUT captured
by JES

→

🖨

PRINT / ARCHIVE

Output printed
or archived

Batch Processing

Jobs are collected, queued, and run without human interaction. Ideal for high-volume, repetitive work: end-of-day settlement, payroll, statement generation, tax processing. A single batch window might process millions of records overnight.

Scheduled or event-driven
High throughput, not latency-sensitive
Classic mainframe "bread and butter"
JES2/JES3 manages the queue

TSO — Time Sharing Option

Introduced with OS/360 in the 1960s, TSO allows multiple users to share the mainframe interactively — each user gets a terminal session. Developers write and debug code, administrators manage datasets and jobs. The ISPF (Interactive System Productivity Facility) green-screen editor running under TSO is still used by thousands of mainframe professionals daily.

Interactive terminal access
ISPF for editing and navigation
REXX and CLIST scripting

CICS — Online Transactions

Customer Information Control System (CICS) is the transaction server that powers online mainframe applications — ATMs, teller systems, reservation systems, point-of-sale. CICS has been in continuous production since 1968. It provides sub-second response times for millions of concurrent transactions with built-in recovery guarantees (ACID).

Launched 1968, still evolving
Manages millions of concurrent sessions
Two-phase commit, XA transactions
VSAM, DB2, IMS data access

// SECTION 05

MIPS & Performance

Millions of Instructions Per Second — six decades of exponential growth

Year	System	Peak MIPS	Growth vs Prior	Era
1964	S/360 Model 30	0.034	baseline	VACUUM / TRANSISTOR
1970	S/370 Model 195	3.5	+102×	INTEGRATED CIRCUIT
1985	IBM 3090	69	+20×	BIPOLAR ECL
2000	IBM z900	3,000	+43×	CMOS 64-BIT
2022	IBM z16	200,000+	+67×	CMOS + AI ACCEL

Note on MIPS: MIPS (Millions of Instructions Per Second) is a simplified measure and does not capture instruction complexity, memory bandwidth, or I/O throughput. Modern mainframe benchmarks use MSU (Million Service Units) for software licensing and SPECint for processor comparison. The z16's 200,000+ MIPS figure reflects a 24-core, 5.2 GHz Telum processor running z/Architecture instructions — roughly equivalent to an entire rack of modern x86 servers, with better RAS (Reliability, Availability, Serviceability) guarantees.

// SECTION 06

COBOL — Still Running the World

Designed in 1959 by Grace Hopper's team. Still processing your money every day.

       IDENTIFICATION DIVISION.
       PROGRAM-ID.  HELLOWORLD.
       AUTHOR.      GRACE HOPPER.

       ENVIRONMENT DIVISION.
       CONFIGURATION SECTION.
       SOURCE-COMPUTER. IBM-Z16.

       DATA DIVISION.
       WORKING-STORAGE SECTION.
       01  WS-MESSAGE  PIC X(30)
                  VALUE 'HELLO, WORLD FROM COBOL!'.
       01  WS-COUNTER  PIC 9(5)
                  VALUE 00000.

       PROCEDURE DIVISION.
       MAIN-PARA.
           PERFORM VARYING WS-COUNTER
               FROM 1 BY 1
               UNTIL WS-COUNTER > 3
               DISPLAY WS-MESSAGE
               DISPLAY '  COUNT: ' WS-COUNTER
           END-PERFORM.
           STOP RUN.

COBOL Structure

IDENTIFICATION DIV.

Program metadata

ENVIRONMENT DIV.

Hardware & file descriptions

DATA DIVISION

All data definitions (like declarations)

PROCEDURE DIV.

The actual program logic

ATM Transactions

95%

Of all ATM transactions worldwide are processed by COBOL programs running on IBM mainframes — the same architecture, in production for 60+ years.

Social Security & Tax

The U.S. Social Security Administration runs COBOL. The IRS processes tax returns on COBOL. Every time you file, your return touches 60-year-old code that has been patched and enhanced but never fully replaced.

Active COBOL Code

800B lines

An estimated 800 billion lines of COBOL are in active production. New COBOL is still written every year — the language was last standardized in 2014 (COBOL-2014) and supports object-oriented extensions.

Airlines & Reservations

SABRE (Semi-Automated Business Research Environment), the airline reservation system co-developed by IBM and American Airlines in 1960, pioneered real-time transaction processing. Descendants of this architecture still power global distribution systems today.

HISTORICAL PHOTOGRAPHS

Gallery — Mainframes & Computing History

From room-sized pioneers to modern z-Series — the machines that ran the world.

ENIAC — University of Pennsylvania

Four ENIAC panels and a function table at Penn, circa 1947. ENIAC filled 1,800 square feet and weighed 30 tons — the first general-purpose electronic computer.

Programming ENIAC, c. 1947

Glen Beck and Betty Snyder programming ENIAC at the Ballistic Research Laboratory — programming meant physically rewiring patch cables and setting switches.

Women Programmers of ENIAC

Betty Jean Jennings and Fran Bilas at the ENIAC control panel. The six original ENIAC programmers were all women, yet went largely unrecognized for decades.

IBM 701 Operator Console

The IBM 701 (1952) — IBM's first commercial scientific mainframe. Operators controlled the machine from this console, monitoring lights and switches for each register.

IBM System/360 Model 30

The IBM System/360 Model 30 (1964) — IBM's $5 billion bet that defined modern computing with its compatible family architecture and 8-bit byte standard.

IBM System/360 Installation

A full IBM System/360 installation — these machines required dedicated air-conditioned rooms, raised floors for cable routing, and teams of operators.

Grace Hopper (1906–1992)

Rear Admiral Grace Hopper — inventor of the first compiler, creator of COBOL, and the person who coined the term "debugging" when she found a moth in a relay.

IBM 80-Column Punch Card

The IBM 80-column Hollerith punch card — how mainframes received input for decades. "Do not fold, spindle, or mutilate" became a countercultural slogan.