7 April marked the 60th anniversary of the mainframe. It was on that day in 1964 that the System/360 was announced and the modern mainframe was born. IBM’s Big Iron, as it came to be called, took a big step ahead of the rest of the BUNCH (Burroughs, UNIVAC, NCR, Control Data Corporation, and Honeywell). The big leap of imagination was to have software that was architecturally compatible across the entire System/360 line.
It was called System/360 to indicate that this new system would handle every need of every user in the business and scientific worlds because it covered all 360 degrees of the compass. That was a triumph for the marketing team because it would have otherwise been called the rather dull System 500. System/360 could emulate IBM’s older 1401 machines, which encouraged customers to upgrade. Famous names among its designers are Gene Amdahl, Bob Evans, Fred Brooks, and Gerrit Blaauw. Gene Amdahl later created a plug-compatible mainframe manufacturing company – Amdahl.
IBM received plenty of orders and the first mainframe was delivered to Globe Exploration Co. in April 1965. Launching and producing the System/360 cost more than $5 billion, making it the largest privately-financed commercial project up to that time. It was a risky enterprise, but one that worked. From 1965 to 1970, IBM’s revenues went up from $3.6 billion to $7.5 billion; and the number of IBM computer systems installed anywhere tripled from 11,000 to 35,000.
The Model 145 was the first IBM computer to have its main memory made entirely of monolithic circuits. It used silicon memory chips, rather than the older magnetic core technology.
In 1970, the System/370 was introduced. The marketing said that the System/360 was for the 1960s; for the 1970s you needed a System/370. All thoughts of compass points had gone by then. IBM’s revenues went up to $75 billion and employee numbers grew from 120,000 to 269,000, and, at times, customers had a two-year wait to get their hands on a new mainframe.
1979 saw the introduction of the 4341, which was 26 times faster than the System/360 Model 30. The 1980s didn’t have a System/380. But in 1990, the System/390 Model 190 was introduced. This was 353 times faster than the System/360 Model 30.
1985 saw the introduction of the Enterprise System/3090, which had over one-million-bit memory chips and came with Thermal Conduction Modules to speed chip-to-chip communication times. Some machines had a Vector Facility, which made them faster. It replaced the ES/3080.
The 1990s weren’t a good time for mainframes. For example, in March 1991, Stewart Alsop stated: “I predict that the last mainframe will be unplugged on March 15, 1996”. Not the most successful prediction, but definitely catching the zeitgeist of the time. As mentioned above, it was the decade of the System/390. We saw the introduction of high-speed fibre optic mainframe channel architecture Enterprise System Connection (ESCON).
The System/360 gave us 24-bit addressing (32-bit architecture) and virtual storage. The System/370 gave us multi-processor support and then extended storage 24-bit/31-bit addressing. With System/390 we got the OS/390 operating system. As we moved into the 2000s, we got zSeries (zArchitecture) and z operating systems giving us 24, 31, and 64-bit addressing. In 2003, the z990 was described as, “the world’s most sophisticated server”. In 2005 we got the zIIP specialty engine. In 2008 it was the z10 EC with high capacity/performance (quad core CPU chip). In 2012, the zEC12 was described as an integrated platform for cloud computing, with integrated OLTP and data warehousing. In 2000 IBM said it would support Linux on the mainframe, and, by 2009, 70 of IBM’s top 100 mainframe customers were estimated to be running Linux. A zEnterprise mainframe could run 100,000 virtual Linux servers.
In terms of operating systems, OS/360 was replaced by MVT, which became OS/VS2 SVS, and then OS/VS2 MVS. That became MVS/SE, which became MVS/SP, which became MVS/XA and then MVS/ESA before becoming OS/390 and finally z/OS.
Mainframes were once islands of technology excellence. Now they are no longer islands, but integrated in so many way. They work happily with cloud environments to ensure the best of cloud can be integrated with the best of mainframes. They run applications that previously were thought of as only found on distributed systems, such as Java and C++. Open-source software, like Zowe, makes it easier for non-mainframers to be able to successfully work on mainframes. The growth of AI on the mainframe makes them even more capable. There’s no problem connecting parts of mainframe applications to parts of applications running on almost any other platform using APIs to create a new and better application program for users.
The z16 and z16 rack-mounted variant mainframes with their Telum chips have on-chip AI accelerators, which delivers latency-optimized inferencing, ideal for mission-critical workloads such as credit card, healthcare, and financial transactions. The z16s are also specifically designed to help protect against near-future threats that might be used to crack today’s encryption technologies.
Mainframes may be misunderstood by people who don’t know them. They may be perceived as antiquated because they are 60 years old. But just like cars and planes over the last 60 years, mainframes have hugely improved over time. And, although they will still run applications written in the 1960s, they can do so much more now. And mainframes are constantly being updated to meet the needs of the time.
I predict that we will be celebrating the mainframe’s 70th birthday and its 80th birthdays, though we may not be able to imagine how compact it will be by then and what new capabilities it will have.
Happy (belated) birthday mainframe.
No comments:
Post a Comment