Comments on: SQL Server

By: Ronnie Doggart

Ronnie Doggart — Thu, 05 Aug 2010 09:34:23 +0000

One think I do like about SQL Server is the Profiler tool for getting traces and seeing what exactly is going on.

By: Jonathan Lewis

Jonathan Lewis — Fri, 04 Jun 2010 08:28:32 +0000

Jerry,
You’ve already discovered that you can play around with rownum to some effect, but it’s much more complicated than SQL Server. It sounds like SQL Server can work along the lines of: “I am aggregating and now I am stopping because I have aggregated too much” whereas Oracle is generally limited by rownum to “I am going to get the stuff I need to aggregate and I will stop if I find too much”. The nature of the Oracle code is restricted to the following type of thing:

select 
	/*+ gather_plan_statistics */
	count(n1)
from
	(select
		n1
	from
		t1
	where
		n1 = 5
	and	rownum <= 300
	)
;

In this case there are 500 rows that match the “n1 = 5″ predicate in the table, but I can stop the query from working once it have found the first 300.

By: Jerry B

Jerry B — Wed, 02 Jun 2010 18:55:15 +0000

One of the nice things that helps TOP is that SQL Server has a Hash Match/Flow Distinct query plan node which is used to eliminate duplicates and which pipelines. (The downside is that the optimizer doesn’t always select that node when it should.) We have a query that checks to see if the number of rows where a person’s last name starts with XXX exceeds a threshold (301 by default). If the threshold is exceeded, then the user is prompted to make the search more restrictive. There’s no way to prevent a user from searching on ‘Smith’. In SQL Server, we see the optimizer generate a Hash Match/Flow Distinct for duplicate elimination. SQL Server does not have the equivalent of STATISTICS_LEVEL = ALL, but we can see by the estimated row counts and the actual # of logical reads that the scan on ‘Smith’ stops when enough rows have been returned by the Hash Match/Flow Distinct. In Oracle, we restrict on rownum we see a HASH UNIQUE for the same query and the row counts from STATISTICS_LEVEL = ALL show that HASH UNIQUE is not pipelining. (All of the rows that start with ‘Smith’ are flowing into the HASH UNIQUE.)

So, do you know of any way to get the desired pipelining in Oracle?

By: I wish .. (1) « Oracle Scratchpad

I wish .. (1) « Oracle Scratchpad — Tue, 01 Jun 2010 19:46:51 +0000

[...] Server — Jonathan Lewis @ 7:45 pm UTC Jun 1,2010 I pointed out some time ago a few of the things in SQL Server that I would like to see in Oracle. Here’s a couple [...]

By: Back from MOW « Oracle Scratchpad

Back from MOW « Oracle Scratchpad — Fri, 30 Apr 2010 08:12:44 +0000

[...] I had still had the presentation I did for Microsoft/Unisys back in February and enough time to remind myself of what was on the slides. On this occasion, though, I also had a [...]

By: SQL Server 2 « Oracle Scratchpad

SQL Server 2 « Oracle Scratchpad — Fri, 05 Mar 2010 09:48:52 +0000

[...] Following on from my posting about the presentation (about Enterprise databases) that I did at a Microsoft event on SQL Server 2008, I’ve just had an article on how to design efficient SQL published on a website that’s [...]

By: Jonathan Lewis

Jonathan Lewis — Wed, 03 Mar 2010 21:54:34 +0000

Zahir,

Sorry, I missed this comment when it first came up.

The problem with adaptive cursor sharing is that it has to go wrong before it goes right – and then it carries on trying to do better, so you can get constantly changing plans.

Also I don’t think adaptive cursor sharing is “persistent”, i.e. the plans and their bind variable ranges don’t get stored in the database (unlike sql base lines), so if a plan gets flushed, or when you restart the instance, you have to start making mistakes and generating new plans all over again.

By: Zahir Mohideen

Zahir Mohideen — Sun, 21 Feb 2010 01:36:55 +0000

Jonathan ,

Wouldn’t “Adaptive Cursor Sharing ” in 11g minimize the diadvantages of bind variable peeking ?
Would you consider “Optimize for” hint be close to Adaptive Cursor sharing ?

Thanks

By: Jonathan Lewis

Jonathan Lewis — Fri, 12 Feb 2010 17:16:02 +0000

Nigel,

That’s the back gate – the one I slip out of under cover of darkness.

By: Nigel

Nigel — Fri, 12 Feb 2010 10:18:02 +0000

Hi Jonathan,

I just saw your house on Google. I like your front gate:

http://gallery.nen.gov.uk/gallery_images/0609/0000/0082/gate_mid.jpg