Sujet : Re: Inside an IBM z17
De : ldo (at) *nospam* nz.invalid (Lawrence D'Oliveiro)
Groupes : comp.miscDate : 02. May 2025, 02:18:07
Autres entêtes
Organisation : A noiseless patient Spider
Message-ID : <vv16ge$3rgrd$9@dont-email.me>
References : 1 2 3
User-Agent : Pan/0.162 (Pokrosvk)
On Thu, 1 May 2025 18:56:39 -0400 (EDT), Scott Dorsey wrote:
It's a weird thing... it's an I/O machine that happens to have a CPU
controlling it.
That’s a mainframe in a nutshell.
It's a very different way of thinking about computing and it's still a
good thing for transaction processing systems with big databases behind
it, the sort of thing that is I/O intensive but cannot be easily
distributed.
It’s an obsolete way of thinking about computing. It dates from a time
when the CPU was considered a scarce, expensive resource, so all these
elaborate “channels” were added to offload as much I/O processing as
possible, to reduce the bother on the CPU.
The result was high I/O throughput, at the expense of high latency when
responding to events, such as user actions. They were a lousy match for
the model of interactive timeshared computing that became popular on the
minicomputers from DEC and other vendors, from about the 1960s onwards.
That new model was seen as “wasteful” by some, but the increase in
productivity of users made it very popular, and the cheaper (non-IBM)
hardware made it more economic.
The logical conclusion was the microprocessor-based single-user PC, which
spent >99% of its CPU cycles waiting for the user to press a key or click
a mouse -- the ultimate in waste, but also the ultimate in productivity.