Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Risk in using sequence number?

Re: Risk in using sequence number?

From: John Darrah <jdarrah_at_veripost.net>
Date: Thu, 21 Feb 2002 23:29:13 +0000 (UTC)
Message-ID: <701b1758b01742f288d9f3ec34abed7f.36240@mygate.mailgate.org>


You don't give enough information to determine whether or not your developer is correct or not.
This developer is most certainly correct if:

1 you have multiple users / processes generating events. In this case even going down to 100ths of seconds still leaves a chance for a race condition to occur.
2 you are not running OPS. There may ba a more meaningfull key to put here but on single instance Oracle, there much fewer "what if" scenerios with sequences.
3 have a rate of transactions that is high enough that it could outpace dbms_utility.get_time. My bet is a well cached sequence will beat the get_time procedure in terms of performance.

Your developer is wrong if:
1 you use OPS. This is a real can of worms any way you look at it. The sequence is no longer reliable at a point in time reference so you can't use it. I imagine sysdate could pose a problem as well due to A) race conditions and B) two nodes not being synched on time 2 your events happen serially at a rate of less then 1 per second.

One thing to consider might be to look at Advanced Queues. I'm pretty sure they guarentee that records will come out in a FIFO order.

Good Luck

P.S. Everyone please feel free to shoot holes in the points I made.    

-- 
Posted via Mailgate.ORG Server - http://www.Mailgate.ORG
Received on Thu Feb 21 2002 - 17:29:13 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US