Posted by Erland Sommarskog on 02/16/06 00:42

Alexander Kuznetsov (AK_TIREDOFSPAM@hotmail.COM) writes:
> When I was solving a similar problem, I did try out the approaches
> suggested by Hugo and Erland. I hate to say that, but I was always
> getting a bottleneck because of lock contention on PinsTable. Maybe I
> was missing something at that time. I had a requirement to produce
> hundreds of PINs per minute at peak times, so I decided to allocate a
> batch of PINs at a time, instead of distrributing them one at a time -
> that took care of lock contention

Bobus said "But, in our particular case, we could have a problem of unused
pins for transactions which rollback."

This would call for a design where you get a start a transaction, get a
pin, use it for whatever purpose, and then commit. But as you say, you
will get contentions on the PINs here, although it's possible that READPAST
hint could help. I did some quick tests, and it seem to work.

The other option as you say is just to grab a PIN or even a bunch of
them. A later job would then examine which PINs that were taken, and
which were never used, and then mark the latter as unused.


--
Erland Sommarskog, SQL Server MVP, esquel@sommarskog.se

Books Online for SQL Server 2005 at
http://www.microsoft.com/technet/prodtechnol/sql/2005/downloads/books.mspx
Books Online for SQL Server 2000 at
http://www.microsoft.com/sql/prodinfo/previousversions/books.mspx

[Back to original message]