Because I was a high school science teacher at the time, I wasn't aware that in 1964 IBM announced the epoch-making System/360 (architected to span the full 360 degrees of the scientific and commercial computing spectrum). Read about the IBM S/360 at Wikipedia and in the IBM archives. Also the quintessential paper Architecture of the IBM System/360. And there's also In the beginning there was IBM as well as The Beginning of I.T. Civilization – IBM’s System/360 Mainframe. A related must-read is Fred Brooks' 1975 classic, released in a 20th anniversary edition: Mythical Man-Month (replete with software engineering insights just as pertinent today as they were back in the 1970s: why, oh why, can't we learn from history?).
I joined IBM Australia in 1970, just in time for the next generation IBM System/370 announcement -- a new generation of hardware and IBM's first large-scale commercial implememtation of virtual storage. (The earlier IBM S/360 Model 67 had a more or less experimental implementation of this, I seem to recall, in its CP/67 operating system.) I was fortunate to have had such an early exposure to IBM mainframes, long a mainstay of organizations all around the globe.
From around the mid-1970s I moved away from the mainframes to some of their mid-range systems (then called "small systems" because minicomputers and personal computers were still a little way off in the future). Firstly the IBM System/7, where I spent a year doing intricate Assembler Language programming on a suite of IBM field-developed programs (FDPs) called PIMS, used to monitor and control Plastic Injection Molding Systems. That sort of work (writing interrupt handlers, etc) was like working on an operating system. It was a most interesting year, but had nothing at all to do with the rest of my career at IBM!
Then I switched to the IBM System/3, announced in late 1969 by IBM development laboratory in Rochester, Minnesota. This lab released a range of other small, easy-to-use systems during the 1970s: the System/32, the System/34 and the System/36, all extremely popular in small organizations and departments or subsidiaries of large ones.
While these systems were being produced, starting around 1972/73 IBM Rochester was working on a radical new system architecture and this was announced in October 1978 as the IBM System/38. The Wikipedia entry for S/38 which puts it very well:
"System/38 and its descendants are unique in being the only existing commercial capability architecture computers. The earlier Plessey 250 was the only other computer with capability architecture ever sold commercially. Additionally, the System/38 and its descendants are the only commercial computers ever to use a machine interface architecture to isolate the application software and most of the operating system from hardware dependencies, including such details as address size and register size. The System/38 also has the distinction of being the first commercially available IBM server to have a RDBMS."I could go on at length about other outstanding aspects of S/38, such as its implementation of "single level storage", and the low-level implementation of database journaling and commitment control, but that's for another time. While the PC world is trumpeting the arrival of 64-bit hardware and operating systems. the S/38 and As/400 and iSeries have had such things (in one form or another) for many years -- and, most importantly, users' applications didn't have to be rewritten or recompiled to take best advantage of new features!
I was heavily involved with supporting, in Australia and Asia/Pacific countries, the System/38 and its successor the AS/400. I took an early retirement option 1992, but this system family has remained my favorite. It's heartening to see the IBM iSeries carrying on the IBM Rochester tradition of highly-advanced yet affordable, reliable, easy-to-use systems in this new century, over thirty years since its architecture was conceived in the early 1970s.
Getting back to the IBM mainframe -- even if you term them "dinosaurs" after more than forty years of evolution they are sleek, powerful, modern beasts with plenty of life left and apparently a resurgence in popularity. The IBM Redbook referred to earlier shows how advanced the zSeries servers are these days. (And you certainly could consider the larger iSeries models truly be powerful mainframes in their own right, despite their small-system heritage.)
I urge you to consider the arguments discussed in THE DINOSAUR MYTH - Why the mainframe is the cheapest solution for most organizations and Is the IBM mainframe a good consolidation platform? and Future of the IBM mainframe looks surprisingly good and The Mainframe is the answer to all your problems before you commit to large difficult-to-manage farms of small systems.
UPDATE: see also IBM unveils the new System z9 - The next evolution in mainframe computing platforms which talks about the crisis of complexity arising from trying to handle growth in application requirements just by adding more and more small servers. It quotes The Yankee Group who report that due to fragile datacenters "many enterprises are restricted in deploying innovative applications that could potentially create competitive advantage." The article mentions that "in a 32-way Parallel Sysplex® cluster, the z9-109 can perform 25 billion transactions a day, compared with 13 billion a day with z990s clustered in a 32-way Sysplex." Now that's some daily transaction rate, isn't it!