If we didn't have caching controllers and we were still on the 370 I/O 
subsystem, would we still be concerned about RPS misses now a days?

I think most disk drives have 512K of cache on them.  My understanding is that 
you don't read/write directly to the disk anymore, but to the drive cache.

And then there is the cache on the RAID adapter to consider.  

Since a RPS miss was that the channel path was busy with some other transfer 
when the block to be transferred came under the heads, true synchronous I/O, do 
we every do, true synchronous I/O any more?

Tom Duerbusch
THD Consulting

Law of Cat Thermodynamics

  Heat flows from a warmer to a cooler body, except in the case
  of a cat, in which case all heat flows to the cat.


>>> "Schuh, Richard" <[EMAIL PROTECTED]> 11/8/2007 3:52 PM >>>
Don't forget that Marty's note mentioned the word "miss" in conjunction
with "RPS" An RPS miss meant that the disk had rotated so that the next
record to be read was past the heads so that there was a wait until it
came back around. That was the performance issue he was talking about,
and it was a big deal in those days. The maximum rotational speed of a
disk was about 3600 rpm. 

 

Regards, 
Richard Schuh 

Reply via email to