The midseventies SHARE survey
SHARE was-is the user group for IBM mainframe shops, and it dates from the OS-360 days. Some of its tapes of shared software mods are preserved at www.cbttape.org and downloadable from there. Tucked on the 1977 tape is a survey of SHARE members, with the raw data in readable form. Browsing this file of 80-column card images has been fascinating for me. Here are some impressions and observations about it, if you're interested.
Members must have been asked (or required) to fill out a survey form when they joined the group. The form is pretty detailed, and seems to e divided into three general sections: contact information, machine configuration, and software in use. There's also a remarks section that many users ignored but some took advantage of. Most interestingly, users were asked to rate the performance of their hardware and software on a scale ranging from unaccceptable to excellent.
The survey covers 496 machine configurations at 343 corporations and government agencies in the United States. The dates of the forms cover the years 1973 to 1977, with most in 1975. The respondents are a major slice of the kind of customers that could afford the big dollars for the big iron of the day: fortune 500 companies and large government agencies, with a smattering of computer services companies thrown in. And a bunch of colleges and universities too, of course.
The midseventies SHARE survey 65
you could also tell the high-end machines ... 165, 168, 3033, etc were designed with OS-VS2 virtual...
The first surprise to me was the number of System-360s still in use. Of the 496 machines, 200 were 360s, even though the 370 was introduced in 1970 and IBM made every possible effort to replace the 360s in the field with 370s. Most of these are model 65 machines, with the second most popular model being 50. There are a few 75s and 67s. The smallest 360 represented is a few model 40s. (Note that SHARE didn't provide support for DOS or DOS-VS, only OS-360 and MVS. Thus the mulbreastude of smaller DOS machines aren't represented here.)
The 370 models seem about evenly split between 155-58 and 165-68, with a small number of 145s and a few really big ones (e.g. 195). The largest memory I saw was 8MB, with the majority of the systems about 2-3MB. A representative configuration would be CBS, with a 370-155 with 3MB of memory, 38 model 3330 (100MB) disks, 24 3330-11 (200MB) disks, and a few "brand-X" disks, probably 2314 (30MB) capacity. They were running OS-360 Release 21.7 (the date on this one is Dec 19, 1974). They list neither tape drives nor communications equipment on the survey, but they must have had some tape drives.
There are very few "true blue" shops. Most have a mix of IBM and non-IBM gear. All processors are IBM of course (Amdahl didn't enter the picture with the first clone until 1976). But frequently the memory is a mix of IBM and other, and often you find disk and tape drives from other companies. Almost all printers are IBM though. Many installations have some terminals and communications gear, it too is a mixture of IBM and others.
The SHARE members seem well satisfied with their IBM hardware. It's rarely rated below "good", except in the case of some 3270 systems and other communcations equipment. The "clones" don't rate so consistently, with their disks often falling below "good" quality. It's odd, though, some shops seem very happy with their Brand-X equipment, rating it good or excellent, while another shop will rate the same gear from the same company only "fair" or even "bad". Users were not asked to rate the quality of their processors or memory, however.
In the area of software, almost all these shops were running some variation of OS-360, frequently OS-MVT but there are more OS-MFT shops than I expected to see. A few are pushing the envelop and running OS-VS1 or OS-VS2 (MVS). Others are using or experimenting with VM-370, mostly as a means to test upgrades, rather than as an operating system. There's a wide variety of software applications listed, mostly IBM products such as TCAM, TSO, and CICS, plus many third-party tools.
IBM doesn't fare as well with their software as they do with their hardware in terms of customer satisfaction. There was no place to rate OS-360 itself, but how users felt about it can be seen from the remarks section. Most users joined SHARE to get user modifications to "improve the performance and reliability of OS". That phrase comes up again and again, with users constantly worried about performance. Security was another major concern poorly addressed by the operating system.
The "Packages" section does have an area to rate performance. In this, HASP is the most well-liked subsystem. CICS rates fairly well too. TCAM is loathed, rarely rated good and often "bad" or "unacceptable". Remarks complain about its frequent crashes or "freezes". TSO is a little better liked, but not much, with users rating it "fair" most often, sometimes "poor" or "good". Few seem satisfied with TSO's performance and many are looking to SHARE to help out with it. There are many third-party packages and subsystems mentioned, with ratings all over the map from excellent to unacceptable.
The remarks section is the most interesting. Many respondents made no remarks, and many offered only a dry sentence or two about hoping to benefit from SHARE membership. But from those that did make remarks, one can get a feel for how the user base really felt about things. The constant concerns are security, reliability, and performance. Many were worried about IBM dropping support for OS-360 (a valid worry as it turned out). Some shops appeared to have problems with vendors pointing fingers at one another.
The midseventies SHARE survey 63
recent posting in comp.arch about personal computers my personal computer was 64kbyte 360-30 ... normally the university shutdown the computing center from 8am sat. until 8am monday. I got a key to the...
The midseventies SHARE survey 67
1977 is pretty early for such user groups, and give a window onto another era. In retrospect it is totally...
The midseventies SHARE survey 68
I'm sure one reason for those 360s hanging around so long when faster-better-cheaper alternatives were available was a result of the same mind-set. A perfect example of how rulesets crafted...
Overall, I'm struck by the conservatism of the user base. When we think about the life cycle of products, we think of the announcement date and the de-commit date. But in reality, the products aren't really in wide use until well after they are announced, their use probably peaks around the de-commit date, and their real-world life extends far past the point where the vendor considers them officially dead.
Another amazing thing is the idea of large organizations using equipment that wouldn't be considered adequate for an email PC for Grandma today to run their entire operations. The most common machine in the survey is probably the 370-158. This was a one-MIPS machine. Most of these boxes had strings of disks adding up to a gigabyte or two, and 2 or 3 MB of main memory. You couldn't give away a PC that small today at a garage sale. And yes, mainframe architecture is I-O optimized and multi-channel and all that, but still I guarantee that OS-MVT running on an emulator on an average PC today will perform many, many times the speed of these original machines in overall throughput not just processor speed. No wonder performance was the number-one concern of all the SHARE members that mentioned their problem areas in the remarks. The demand for computing services vastly exceeded the supply, regardless of the money spent on it.
The midseventies SHARE survey 64
Lots of snippage below is Founded in 1955 as a self-help organization of scientific and education customers (recall that in those days there was...
The midseventies SHARE survey 66
i have some recollection of being in pok machine room (705-706?) 3rd shift and working around ludlow(?) on a 360-67 ... he was building the aos2 (prototype for os-vs2 ... svs) ... basically taking...
It would really be interesting if other user groups, say DECUS for example, ever did surveys of their users' configurations and satisfaction, and if those numbers could ever be found. Interesting, indeed.