| > Does this mean maximum records which one duplicate KEY has is limited?
NO. (I almost wished it did though :-).
Before you hit the maximum 16 bit size, you'll hit the absolute
maximum bucket size of 63 blocks = 32KB. In practice, the
bucket size for alternate keys it often choosen to be 12 blocks or so.
This allows for roughly 1000 dups per bucket. Once a bucket fills up,
RMS will just create a next bucket repeating and continuing the same key.
A customer I visited two weeks ago had 5.5 million record in a file
with 1.5 million having the same alternate key. They had 1608 buckets
in the duplicate chain for that key value. Adding a single record with
that key value to the file took 20 seconds on a good days, 2 minutes
on a bad day. :-(. We flipped the 'null key value' bit on the file
(through FDL ofcourse), converted and a job take used to take 36 hours
finished in 7 hours and has the potential to finish in 2 hours.
[Because it took so long, they sorted the added records by postal code
to allow some early exctract to be done whereas the could have
presorted by primary key order making RMS caches much more effective]
You may want to use my SIDR program on the VMS freeware CD or from
topic 5 in this conference to analyze an existing file with dups
(try your own VMSmail file for starters)
hth,
Hein.
|