RE: File Reorg - Cons

[Evan Harris <spanner@xxxxxxxxxx>]

Hi Mike

just to give you even more to think about :)

I was once lucky enough to have a 140 million record monster file with a 
700+ character record length and 25 logicals built over it that had 20 
million plus deleted records which had to be re-organised. (makes me sound 
like a Yorkshire man seeing it written down like that...)

In any case, if time is at a premium some of the things I found that can 
help a great deal are to:

- Drop the logical files before doing the re-org then manually create them 
afterwards

- You can run two or more rebuilds through submitting CRTLF commands to run 
concurrently; normally RGZPFM seemed to/used just build them in an 
arbitrary order I could never figure out but looked to be original creation 
order.

- It is possible to somewhat optimise the rebuild sequence which may save 
disk and will almost certainly save time. Analyze the keys on your logical 
files to determine which have the most complex keys or keys that can be 
used to assist in building subsequent access paths (some experimentation 
may be in order here). If you optimise the order from most complex to least 
complex and arrange them in job streams appropriate for the keys (e.g. one 
job stream for customer number/etc keys, another for customer name/etc 
keys) you may save some time.

- If you are really cramped for time write a program to write the records 
out to a new file. This also provides the option to run parallel job 
streams of the read/write operations by selecting records to be added into 
the file to be replaced by relative record number ranges. When the writes 
have completed delete the old file (if you dare :) and rename the new as 
the old and start building your indexes. This can significantly speed 
things up :)

Hope this helps

Regards
Evan Harris


> > On Behalf Of Mike Berman
> > Subject: File Reorg - Cons
> >
> > I noted alot of suspicion when I attempted to set up a file reorg process.
> > Just getting lists of those with large amounts of deleted records. Is
> > there some possible occurance that would be a detriment? I am of course
> > referring to off hours running.