Thanks for all of the posts, guys!

Yes, there are a ton of inserts and deletes.

And while it deletes records over five mins old, due to poor design, it runs Waaaay...

Ah, ok! Here it is...

This was Prior....the PK is non-clustered (I'm guessing because this database (third-party) was upgraded through the years <shrug> ).

The execution plan is Prior as well.

The premise being, indexes ignore when you only have a few rows, and it all fits on a page.

Even if it was ignored...such a small number of rows... :blink:

Table:

CREATE TABLE [dbo].[RainbowQueue](

[RainbowType] [varchar](50) NOT NULL,

[RainbowKey] [varchar](255) NOT NULL,

[WorkDay] [datetime] NOT NULL,

[UpdatedRID] [bigint] NULL,

[SessionId] [int] NULL,

[BranchId] [varchar](20) NOT NULL,

[RecordUpdated] [datetime] NULL,

[RecordUpdatedBy] [varchar](20) NULL,

[RecordUpdateId] [timestamp] NULL,

CONSTRAINT [PK_RainbowQueue] PRIMARY KEY NONCLUSTERED...

Yes, it is a heap...

Excellent, excellent article, Gianluca!

-Grubb

Thank you, Gentlemen! Very helpful!

-G

Mike John (6/8/2011)

---

I would say identical for the "locate what to update" part. but then obviously a slight overhead incurred due to the need to update the index(es) if the...

++ to majorbloodnock

I think the author doesn't give enough respect to the compounding.

The compounding kicks in quickly. Percentages affect everything down the line, including other bonuses or profit-sharing that...

Grubb (4/8/2011)

---

Guys, after I create, run, stop, and delete the trace, I cannot ever seem to open the resulting file with Profiler: "Access denied". I'm assuming something still...

Guys, after I create, run, stop, and delete the trace, I cannot ever seem to open the resulting file with Profiler: "Access denied". I'm assuming something still has...

From the link Ninja posted:

35% wait seems like it may be significant.

Is...

TheSQLGuru (4/5/2011)

2) it is likely that you can observe what is going on by evaluating aggregate waits stats. Take a snapshot of waits, do a waitfor for 3 mins,...

Thanks, Craig. The problem with that is that the query only runs ~400 ms. Do you know of a way...