[Info-vax] ISAM on disk layout. Was: Re: HP wins Oracle Itanium case

[glen herrmannsfeldt]

JF Mezei <jfmezei.spamnot at vaxination.ca> wrote:
> Jan-Erik Soderholm wrote:

>> Even more so in todays SAN's where you get a number of "gigs"
>> out of some undefined disk array. Placement of files, as in
>> MVS and other older systems, is mostly an non-issue today,
>> from a performance standpoint. The fact that MVS still needs
>> BLOCKSIZE, LRECL and so on, is mostly a compatibility issue.


> As I recall, blocksize and LRECL has more to do with IBM's equivalent to
> RMS than to disk organisation for DASD (hard disks).  When a COBOL
> program opens a disk dataset, the blocksize probably discates size of
> buffers and how many physical blocks are to be read for each logical
> read operation etc.

For S/360 and S/370 disks, the BLKSIZE is the physical block size
on disk. IBM CKD disks write blocks pretty much the same way
they write tapes. 

Later disks emulate the blocking while writing fixed sized
blocks internally. Still, at the channel interface the BLKSIZE
blocks are there.

With some BLKSIZE values, this can be somewhat more efficient
than working with FB512 disks. 

Especially in the DOS/360 days, the ability to write smaller
blocks was significant. For systems with 4K or 8K total, blocking
at 80 allowed for very small buffers.

> However, when writing/reading to tape, tghe blocksize does matter
> because mag tapes still let you specify a physical block size 
> (as is the base for VMS backup BTW).

-- glen