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 amazing what they got out of the cpu cycles, and how much they paid for them. And, how few they were.

This seems pretty large for the time. Amazing number of spindles.

It must have been somewhat of a hazzle to keep all these (50+) devices operating all the time, even with IBM service.

Be careful what you ask for!

IBM has a reputation for making solid hardware.

No wonder. 1-3 mips, 10G disk on 50+ spindles (why didn't anyone think of raids sooner?) and 3-8 MB memory. No wonder they cared deeply about performance when this served a whole company.

This is where I have been burnt again and again.

All the systems around me regularly run software for decades, and computers should ideally run for at least a decade.

No vendor has been able to cope with this, except the user-provided open-source systems from the people in the trenches.

I still have a workalike of the transaction system I worked with in the mid-1980s stored somewhere, together with a decade of data.

A system has to be run for at least a two-digit number of months before it can be trusted with important stuff. A rollout takes a few months, so we are up to a year of walltime after deliveries before systems can be used on real stuff.

As a ballpark estimate , the TOTAL number (~500) of sites would be around 400-800 MIPS, 1-3 T of disk, and a handful of gigabytes ram.

The shocker is that I have more capacity than this at home. Even a slowish VIA processor at 533 Mhz beats the entire SHARE membership of 1977.

In 1977 the black helicopters would be homing in to a non-authorized suprtcomputing site.

I haven't seen any statistics from before the VAX age from DEC.

One point was that customers were far more in control of their DEC mainframes, and ran lots of their own stuff, or other non-dec stuff on them.

-- mrr