Re: I/O count differences

[Jim Oberholtzer <midrangel@xxxxxxxxxxxxxxxxx>]

More importantly than what the program is reading and writing is what's 
being discussed here are logical read/write operations, not physical 
ones.  The I/O count RPGLIST is looking at is very likely physical I/O 
not logical.  You can use the performance tools to get both physical and 
logical I/O but not in the way I suspect is being used now.   When the 
system gets ready to write to disk, it will be page at a time, not 
record/row at a time.

Jim Oberholtzer
Chief Technical Architect
Agile Technology Architects


On 1/23/2013 11:43 AM, Dan Kimmel wrote:
> My point is that it isn't "spinning through the disk". Probably never touches disk at all. The index is most likely entirely contained in a memory page or two. That index is maintained within memory as a set of vectors in trees. The system need only make a few thunks through the vector map within memory to see that there is no index that matches your key.
>
> -----Original Message-----
> From:midrange-l-bounces@xxxxxxxxxxxx  [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of RPGLIST
> Sent: Wednesday, January 23, 2013 10:41 AM
> To: Midrange Systems Technical Discussion
> Subject: RE: I/O count differences
>
> I undestand that, I guess what I'm trying to determine how to evaluate the fact that its still taking resources and time (as minute as it may be) to spin through the disk.
>
>
> > >  The system uses the index only on a no-hit. Never touches the data at all.
> > >
> > >  -----Original Message-----
> > >  From:midrange-l-bounces@xxxxxxxxxxxx
> > >  [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of RPGLIST
> > >  Sent: Wednesday, January 23, 2013 10:23 AM
> > >  To:midrange-l@xxxxxxxxxxxx
> > >  Subject: I/O count differences
> > >
> > >  I need an explanation on this if someone would be so kind.
> > >
> > >  When I chain to a file with a key and get a hit, I fully expect the
> > >  I/O count to increase by 1, even though it is sort of doing a SETLL
> > >  and then a READE within the chain process.  However, whenever it
> > >  doesn't find a record it shows no I/O count.
> > >
> > >  I understand that since it didn't find a record it brough nothing back
> > >  from the disk and therefore didn't increase the I/O count, but it
> > >  certainly used resources to spin all the way through that drive
> > >  disk....Is there anywhere that the system accoutns for this in the statistics?
> > >
> > >  I see the same results when performing a SETLL, If %Equal, actually
> > >  even if there is a hit I don't see an I/O count until the READE statement.
> > >
> > >
> > >
> > >  --