On May 1, 2014, at 3:14 PM, Ron Hawkins <[email protected]> wrote:

> Shmuel's reply prompts me to suggest that you consider what lies outside the 
> nine dots.
> 
> Firstly, would you consider thin provisioning of the installation volumes and 
> SYSRES? This will get you onto 3390-A for these volumes using any size you 
> want. Why is this good? Because DVE is also good. If you start out with 32K 
> Cyls and later find it is not enough, just make the volume bigger. You need 
> 40K Cyls, then just add 8K Cyls to the existing volumes. Sounds easy, but I'm 
> not a software jockey.
> 
> Second thing is ignore the previous paragraph and just allocate full sized 
> 3390-A volumes with 262K Cyls. If you don't use the space on the volume then 
> you don't use space in the disk array, and I think it will be some time 
> before Serverpac manages to fill one of these pupies. All those that worked 
> with RVA will remember over provisioning with 3390-3 when you only needed 
> 3390-1 for your JES2 Checkpoint.  This is the same thing.
> 
> Finally,  make all the data sets ten (yes 10) times larger than they need to 
> be, with ten times the directory blocks you think you want. It is the same 
> principle as above: if you don't use the space on the volume then you don't 
> use the space in the disk array,
> 
> OK, there is a sanity check on the the space you use in the array because of 
> the page allocation boundaries. IBM uses 1GB pages, HDS uses 32MB pages, and 
> I think EMC use 14MB pages. So an empty volume with just a VTOC & VVDS will 
> consume 1GB, 32MB, and 14MB respectively from the pool of each vendor. For 
> EMC and HDS the fragmentation generated by the Cylinder allocation area with 
> just a VTOC , and the empty space in datasets is fairly small and does not 
> dramatically inflat the pool usage (remember NCL?), but for IBM the 1GB page 
> size may need some monitoring.
> 
> Embrace the future of z, because it is already here. Now you guys that know 
> better can shoot me down: shields are up :-)

This goes along with my response to this thread: people are still piddling 
around with mod 9’s and even mod 3’s? Once dynamic pooling became available on 
our Hitachi VSP I reconfigured all our volumes to mod 54’s. I’ve provisioned 
volumes with 120% of the array’s physical capacity, but only 60% of that 
physical capacity is actually in use, so I’m good. I have fewer volumes to 
worry about, and I can let the storage controller worry about performance and 
efficient use of space.

It’s the 21st century; let’s live like it.

-- 
Curtis Pew ([email protected])
ITS Systems Core
The University of Texas at Austin

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to