I just did a quick experiment:

3390-3 disk image formatted at 1K blocks, ext2 fs 1K
blocks. df showed 1.6 GB capactiy

3390-3 disk image formatted at 4K blocks, ext2 fs 4k
blocks.  df showed 2.3 GB capacity.

The drop in total device capacity was about 27%.

I was not able to mount a disk formatted at 4k blocks
with an ext2 fs of 1k blocks or an reiser fs of 1k
blocks. Has anyone been able to to this?

The reason is there is so much drop in capacity is
that IBM settled on the 3390 image with a high
overhead for small blocks. Even with the raid devices
(RAMAC, RVA, SHARK, EMC, etc), the track is recomposed
into a full eckd image in the cache and I/O is from
the cache image. 4K makes a lot of sense when your
paging is 4k blocks and most of your data is VSAM (4K
blocks).

As long as you have to use a 3390 track image with
ECKD mapping, the smaller tracksize will cost you
capacity on the image. Probably only the RVA will see
equivalent savings on the backend. Other devices do a
one to one mapping of the full ECKD image.

FBA would be a better solution. Does the new SCSI
support for shark allow 1k block sizes?

I can see some other issues with 1k blocks. What would
happen if you make the swap (page) packs 1K? Would it
even work. With files over 4k, there would also be a
loss in I/O efficiency - 4 times as much I/O would be
required. And this I/O efficiency was also translate
to reduced efficiencies for reading the node tables.
With smaller node tables, there may be more I/O
activity locating a file.

=====
Jim Sibley
RHCT, Implementor of Linux on zSeries

"Computer are useless.They can only give answers." Pablo Picasso

__________________________________
Do you Yahoo!?
Yahoo! Mail SpamGuard - Read only the mail you want.
http://antispam.yahoo.com/tools

Reply via email to