Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

Data center house of horrors: The mainframe that wouldn't die

In this seventh slide of the data center house of horrors slideshow, pundits keep pushing on the fact that the mainframe just won't die.


The mainframe that wouldn't die

For decades, pundits have inaccurately predicted the mainframe’s demise. But the following story is about the mainframe they couldn’t keep alive.

“When I was at Cornell around ’74, we had just gotten the latest 370 for our data center,” said Wayne Kernochan. “But a strange thing kept happening: Suddenly -- after a few hours, and for no apparent reason -- the 370 would stop working.

"IBM sent a techie, but he ran into a brick wall. No crash was like any of the others; there were no hardware failures, there was no software cause. It was just that, after a few hours, the computer would stop working.

"IBM pulled out all the stops, and brought in the real experts from their Colorado facility. After endless tests, they suddenly noticed something strange: Every crash happened right after one microcode instruction. Usually it worked fine, but every crash followed that one instruction.

"Further investigation revealed that the instruction had been slightly altered from IBM's specs. Even further investigation revealed that a Cornell engineer had figured out a change to make that instruction -- and therefore the 370 -- run a bit faster. But every so often, a second instruction would follow, which the system assumed would take longer, and thus was enough to cause a cascade of miscomputations ending in a complete crash.”


1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 

NEXT >>  


Dig Deeper on Data Center jobs and staffing and professional development

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.