Something that hasn't been mentioned in this thread is that SHARK dasd is
significantly faster than real 3390 spinning disks. When we replaced RAMAC
dasd with SHARK, the clock time to do something was several times less
under the SHARK than the RAMAC. I never replaced real 3390's with a RAMAC,
but I suspect that the RAMAC was faster and probably by quite a bit.
This is all in line with the observations that Lynn Wheeler makes and has
made about the lag in the increase in DASD speed vs. CPU speed, but the
SHARK is still extremely fast compared to a real 3390. The fact of the
matter is that 3390-3 sizes have become the mainframe industry standard but
the only thing that is the same between SHARK and real 3390's is the size
and (perceived) disk architecture. I'll bet that you could get more I/O's
per second out of a SHARK 3390-27 address than 9 3390-3 addresses.
Bill, how about an "it depends" explanation or clarification.
Jim
At 02:19 PM 10/18/2005, you wrote:
There isn't much of a problem per say...
If you have a mod-3, you can do xxx amount of I/Os per second before
hitting a performance problem.
If you have 3 mod-3, you can do 3*xxx amount of I/Os per second.
However a mod-9 can only do about xxx amount of I/Os per second.
Jim Bohnsack
Cornell Univ.
(607) 255-1760
