SDB was invented in the late 1980s when RAID was starting to become more than a 
laboratory curiosity and SLEDs were still the only kind of real disk drive 
available.  IBM is not going to change the  code that calculates SDB to support 
the fact that the CKD architecture track no longer exists in any real disk 
drives.  This code change would be compatible with reality but not with all 
their customers' software that assumes CKD architecture still exists for real 
disks.  Each vendor that makes RAID to emulate SLEDs is free to map the CKD 
architecture being emulated on real RAID any way it wants to, and is also free 
to keep the details of the mapping a proprietary trade secret. 
Bill Fairchild 
Franklin, TN 

----- Original Message -----

From: "Paul Gilmartin" <[email protected]> 
To: [email protected] 
Sent: Saturday, December 7, 2013 11:49:30 AM 
Subject: Re: Something to Think About - Optimal PDS Blocking 

On Sat, 7 Dec 2013 01:56:56 +0000, DASDBILL2 wrote: 

>We can calculate the wasted space given the block size, average number of 
>blocks per track, average number of this, that, and the other thing, but 
>wasted space is meaningless and unknowable unless you are using a SLED 3390 
>which means now at least 15, maybe 20-year old disk drives.  With the advent 
>of RAID, the interrecord gaps have been virtualized. 
>  
If gaps and other unused space are virtualized, there is no reason for SDB 
ever to choose a BLKSIZE other than 32760 (or nearest multiple of LRECL). 
I suspect there are various implementations of RAID: some may virtualize 
unused space; others keep images of entire tracks, however sparsely used. 

I tried the experiment I suggested: I created 10 mebers of 350 records. 
With RECFM=FB,LRECL=80,BLKSIZE=27920 (chosen by SDB) the data 
set used 10 tracks.  When I forced to BLKSIZE=24000, it used 7 tracks. 

Admittedly a deliberate worst case. 

-- gil 

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


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

Reply via email to