Thanks for the suggestions everybody.

I've rewritten the logging system to use an identity column and removed the code that identifies probable repeat records when inserting into the log.

I still...

From what I've read, there's an overhead associated with checking for locks that you incur even if there aren't any.

Adding NOLOCKS is supposed to skip that.

I meant that the data is updated so infrequently that I dont think the costs of record...

I'm not concerned about data accuracy in this system, so the NOLOCKs shouldnt be a problem. The main data is updated at infrequent intervals, and its not critical that it...

In terms of software, none.

They're both windows 2003 servers, sql2005 sp2 etc

Hardware is completely different, the development "server" is a high-spec PC, with nowhere near the power of the quadcore...

Sure.

The log table looks like this;

CREATE TABLE [dbo].[EventLog](

[Id] [int] NOT NULL,

[EventType] [varchar](10) NULL,

[UserId] [int] NULL,

[EventDate] [datetime] NULL,

[AdditionalInfo] [varchar](200) NULL,

[MediaId] [int] NULL,

[TransactionId] [int] NULL,

[OriginalMediaId] [int] NULL,

[OtherTableName] [varchar](50) NULL,

[OtherTableId] [int] NULL,

[IPAddress] [varchar](15)...

So "yyyy-mm-dd hh:mm:ss" isnt universally recognised, but "yyyymmdd hh:mm:ss.nnn" is?

Weird

Also, when you say refer to the users language, I take it you mean the SQL Server user? I've looked at...