winscape 2374

Not necessarily reliability..definitely agility. The reason schedulers were invented was so that two tasks could be handled on the same machine. Once you've started scheduling, you find out that some tasks never get run because it can't fit. So shuffling is invented. That allows more than two tasks to go into the task queue. As this queue gets longer, you discover that the system spends all its time shuffling. Then you invent swapping (requires disks). Now you can service all tasks. But some are more important than others. Now you invent all kinds of parameters to decide who is to be picked next. Now that all this management of servicing users has been done, users start to use the system for more and larger tasks. Now you need to invent some way to address code and data that has become too large almost over night.

And that, boys and girls, is the computing biz' cycle. Now add the extra ingredient of moving bits from here to there.

You forget that c is slow. A human user doesn't need a millisecond response time. A human user only has to have an indication that his-her input has been received, stored, and queued. This is not Real Time (as I define real time).

BAH