Re: New 270.

Andy,
I think to remember that the system is worse than you described:
mixing 8 GB and 17 GB drives results in half the percentage on the 17 GB
compared to the 8 GB drives, because an arm is an arm and the controller
does not know about the capacity "behind" this arm.
So, if you you got a system with, for instance, 46 % used space on an
80 GB ASP consisting of 4x 8 GB (RAID5) and 4x 17 GB (RAID5)
then the drive sets would look something like:  70 % space occupied on
the 8 GB and 35 % used space on the 17 GB drives.
This is only different when running ASP balancing or doing bulk
unloads / reloads for large amounts of data.
Generally speaking, I think one can much more easily overload a certain
disk arm with drives of higher capacity than with a bunch of smaller drives.
I would certainly be interested in the exact algorhythms behind these
symptoms I watched on real systems out there.

Regards from germany, Philipp Rusch

Andy Nolen-Parkhouse schrieb:

> John,
>
> My experience has been that the OS writes new data to the drive with the
> least per cent utilization, until such time as all of the drives in that
> ASP are equalized.  Because of this, you need to be careful with empty
> drives.  If all new writes were random adds to DB files, then perhaps
> there were not be much impact.  But suppose you added a new 8.5 disk
> drive to a system at 80 per cent disk utilization.  Then you restore a 6
> GB library with full of database files.  All of those database files
> will be restored to that one drive and any application using that
> database would beat the living daylights out of that single drive.  The
> same thing would occur if you did a copy of your largest database file;
> it would all end up on that one drive.
>
> If you had a mixture of 8.5 GB and 17 GB drives, then the system would
> spread the data until all drives were at an equal per cent storage
> utilization.  So you could expect your 17 GB drives to have twice as
> much stuff as the 8.5 GB drives and thus twice the activity (assuming
> the scatter loading is effective).  This would be the reasoning for not
> having an ASP with a wide discrepancy in disk sizes.
>
> The ASPBAL commands are pretty much required when adding new drives to
> an ASP to avoid the activity imbalance.  Purists (or traditionalists)
> may still do a system reload under these circumstances.
>
> I've never seen an official explanation from IBM on their scatterloading
> methodology.  I hope someone can find one.
>
> Regards,
> Andy Nolen-Parkhouse
>
> > -----Original Message-----
> > From: midrange-l-admin@midrange.com
> [mailto:midrange-l-admin@midrange.com]
> > On Behalf Of John Ross
> > Sent: Monday, June 03, 2002 10:56 AM
> > To: midrange-l@midrange.com
> > Subject: RE: New 270.
> >
> > I was told last week by a business partner, not to go over double the
> size
> > of your smallest dive in your system, so if you have a 4 do not go
> over an
> > 8.  He tried to explain why but I really did not understand it.
> Something
> > about it would get more reads on the smaller drive.
> >
> > Would it matter how you let the system spread the data? If you just
> put
> > another drive in and do not force it to spread the data on the system,
> it
> > seams like the new data would be put on the new drive and might get
> more
> > reads then older data, at least for a while. If you are letting the
> system
> > spread the data (Not forcing it) does it only put writes on the new
> drive
> > or will it force data that it should only be reading to be moved. Does
> the
> > OS try to stay on the first drive(s)? (I think in my 38 days I heard
> it
> > did) IF so with 8's it might spread over 2 drives with 17's it could
> all
> > end up on the first drive. And it seams the OS would be used a lot. I
> have
> > used WRKDSKSTS for 15 min and the number 1 disk is 1st or second for
> > percent busy, but only by 1%.
> >
> > Is there a link anywhere that explains how IBM spreads the Data?
> >
> > John Ross
> > www.opensource400.org
> >
> >
> > > > Subject: Re: New 270.
> > > >
> > > > Last year, I made the move from 6- 8 GB drives to 6-17 GB, and
> > >performance
> > > > suffered considerably.
> > > >
> > > > Al
> >
> > _______________________________________________
> > This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing
> > list
> > To post a message email: MIDRANGE-L@midrange.com
> > To subscribe, unsubscribe, or change list options,
> > visit: http://lists.midrange.com/cgi-bin/listinfo/midrange-l
> > or email: MIDRANGE-L-request@midrange.com
> > Before posting, please take a moment to review the archives
> > at http://archive.midrange.com/midrange-l.
>
> _______________________________________________
> This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
> To post a message email: MIDRANGE-L@midrange.com
> To subscribe, unsubscribe, or change list options,
> visit: http://lists.midrange.com/cgi-bin/listinfo/midrange-l
> or email: MIDRANGE-L-request@midrange.com
> Before posting, please take a moment to review the archives
> at http://archive.midrange.com/midrange-l.