Vic Cross wrote:
> I think an example is required. :)
Some time ago I had a hogged-out (with DASD) RS/6000-F50
that I was using as a file server. With three filled
sixpaks on three different SCSI buses, I set up some of
the VGs vertically, striping across the SCSI controllers.
This was done for speed, allowing 3 I/Os to be literally
happening at the same time. With a little tuning to AIX
it was easy to make sequential/consecutive I/Os to that
striped filesystem *FAST*. I think it hurt me later to
discover that I wasn't anywhere near needing that speed,
but, hey, I got practice at this exercise.
Funny thing is that I eventually had to expand the LV in
my main repository VG... and it had to be expanded in the
"right" factor. With the data striped across the six
drives (two in each sixpak, and getting the drives first
recognized in the "right order" can be entertaining to
think about) I had to do my extensions across the board.
Now, in AIX, a JFS (or large file enabled JFS) needs a
journal. At the time the journal was in another LV, so
one of the granules (in AIX parlance a "partition", but
that nomenclature conflicts with all non-AIX versions of
the word, so I tend to use the term "granule" since it's
the smalled allocatable granularity in a volume group,
just to keep from confusing my PC indoctrinated co-workers
from being too ultimately confused) is set aside as the
journal on ONE of the drives within the volume group...
So extending a file has to be done SIX of these "granules"
at a time... and there came a time when the LV had used
up all of the allocatable space in the volume group
except for 5 free "granules" ("physical partitions")
which could not be allocated because I needed six of 'em.
Note that this is a *general* issue.
On to some other smarmy comments I can (given my current
ignorance of the realities, having no experience) make
about SANs: When you have total control of the drives
and connectivity between them and the system, tuning is
something you have CONTROL over and so you can tune the
I/O system to your heart's content.
Given my ignorance of what can be done on a SAN device,
it seems to me that there is a performance disconnect;
the "Big eFfing Tank of Dasd" may be tunable, but is it
tunable to the degree that you need? On AIX the intra-
and inter- policies provide a degree of tuning when
allocating (and extending) LVs, but this is when you
are able to "feel" a real spinning disk. Most of these
efforts seem to be irrelevant at the host level once you
push the job of managing the spinning DASD to a SAN...
...but, in AIX, it sure seemed like the system itself
adjusted some strategies depending upon how the LVs
were arranged on DASD.
So there you have it-- I've made wildly insane remarks
about SAN devices because I've never seen, touched or
even smelled one... and I hope that those correcting
my misperceptions will do so and therefore educate me.
I just hope that using the AIX LVM similes didn't
cause any major brain aneurysms out there...
--------------------
John R. Campbell, Speaker to Machines (GNUrd) (813) 356-5322 (t/l 697)
Adsumo ergo raptus sum
MacOS X: Because making Unix user-friendly was easier than debugging
Windows.
Red Hat Certified Engineer (#803004680310286)
IBM Certified: IBM AIX 4.3 System Administration, System Support
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
