Re: Keylocks and updates
From: Max Spectrum (anonymous_at_discussions.microsoft.com)
Date: 03/15/04
- Next message: Oscar: "Re: SQL Server 2K connection is interrupted regularly, why ?"
- Previous message: Tim S: "Re: using 'sp_' as a naming convention for stored procedures"
- In reply to: Ray Higdon: "Re: Keylocks and updates"
- Messages sorted by: [ date ] [ thread ]
Date: Sun, 14 Mar 2004 18:06:08 -0800
----- Ray Higdon wrote: -----
First off if you are being inserted into that heavily then a composite
clustered index is NOT the way to go. This can cause unpredictable page
splitting.
**** No inserts - just lots of updates. Does page splitting occur on an update?
I would have a clustered index on a column with the IDENTITY
property, then use a unique constraint on the columns to make a primary key.
**** clustered index does exist on the PK and it is an identity column
And so you know, keylocks are not necessarily bad.
**** I agree. Key locks are not necessarily bad, but in my case it is affecting performance and limiting the frequency that a row can be updated.
--
Ray Higdon MCSE, MCDBA, CCNA
---
"Max Spectrum" <anonymous@discussions.microsoft.com> wrote in message
news:D5983173-633F-49E8-8358-F3F7D5DF713F@microsoft.com...
> Scenario: A table that is 6 columns wide, has a primary key with a
clustered index. One of the columns is incremented approx. 250-300
times/minute. At this rate key locks and some general performance
degradation occurs on the server. Anything higher causes performance to drop
dramtically.
>> Any suggestions as to how one might increase the rate that a single record
can be updated? I need to get it up to 2500+ /minute.
- Next message: Oscar: "Re: SQL Server 2K connection is interrupted regularly, why ?"
- Previous message: Tim S: "Re: using 'sp_' as a naming convention for stored procedures"
- In reply to: Ray Higdon: "Re: Keylocks and updates"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|