How do I measure the cost of compiling an execution plan?

18 May 2005 12:28 AM

pshroads

My company is using a lot of ad hoc queries and I am trying to convince
them to move to stored procedures. However certain managers don't
believe that the compilation of ad hoc queries and the fact that they
usually aren't reused is really anything to worry about. They think
that the cost of doing this is minimal and instead want to focus on
tuning bad queries.

I of course agree that badly written queries should be tuned but it's
also a basic tenet of databases users that stored procedures should
always be used where possible to take advantage of precompilation and
plan caching. However I don't know how to actually prove that with
data.

How can I measure what the overhead is for compiling and building an
execution plan for an ad hoc query vs. the putting the query in a
stored procedure? I've looked at the graphical execution plan for an ad
hoc query and a stored procedure and they look the same.

Thanks

18 May 2005 12:37 AM

Kalen Delaney

SET STATISTICS TIME ON

The first set of data before the query is executed will be the time for
parse and compile. The stats after the data is returned will be the time to
execute.

If the first set of data shows 0, it means the plan is being reused, and
note that SQL Server 2000 can reuse plans for queries other than stored
procedures.

--
HTH
----------------
Kalen Delaney
SQL Server MVP
www.SolidQualityLearning.com

<pshro***@gmail.com> wrote in message

Show quoteHide quote

news:1116376117.157139.259730@g44g2000cwa.googlegroups.com...
> My company is using a lot of ad hoc queries and I am trying to convince
> them to move to stored procedures. However certain managers don't
> believe that the compilation of ad hoc queries and the fact that they
> usually aren't reused is really anything to worry about. They think
> that the cost of doing this is minimal and instead want to focus on
> tuning bad queries.
>
> I of course agree that badly written queries should be tuned but it's
> also a basic tenet of databases users that stored procedures should
> always be used where possible to take advantage of precompilation and
> plan caching. However I don't know how to actually prove that with
> data.
>
> How can I measure what the overhead is for compiling and building an
> execution plan for an ad hoc query vs. the putting the query in a
> stored procedure? I've looked at the graphical execution plan for an ad
> hoc query and a stored procedure and they look the same.
>
> Thanks
>

18 May 2005 3:16 AM

Andrew J. Kelly

In addition to that you need to consider how much memory all those adhoc
plans are taking up in the procedure cache. I have no idea what yours is
like but I have seen systems with all adhoc queries use well over 1GB for
procedure cache on 32 bit systems and larger on 64 bit. That memory can
better be utilized for data and index caching instead of useless plans.

--
Andrew J. Kelly SQL MVP

<pshro***@gmail.com> wrote in message

Show quoteHide quote

news:1116376117.157139.259730@g44g2000cwa.googlegroups.com...
> My company is using a lot of ad hoc queries and I am trying to convince
> them to move to stored procedures. However certain managers don't
> believe that the compilation of ad hoc queries and the fact that they
> usually aren't reused is really anything to worry about. They think
> that the cost of doing this is minimal and instead want to focus on
> tuning bad queries.
>
> I of course agree that badly written queries should be tuned but it's
> also a basic tenet of databases users that stored procedures should
> always be used where possible to take advantage of precompilation and
> plan caching. However I don't know how to actually prove that with
> data.
>
> How can I measure what the overhead is for compiling and building an
> execution plan for an ad hoc query vs. the putting the query in a
> stored procedure? I've looked at the graphical execution plan for an ad
> hoc query and a stored procedure and they look the same.
>
> Thanks
>

18 May 2005 3:47 AM

pshroads

Thanks for your reply. Would I look at the size of syscacheobjects with
sp_spaceused to determine the amount of memory that the procedure cache
is using?

18 May 2005 7:27 AM

Tibor Karaszi

Use DBCC MEMORYSTATUS for that:
http://support.microsoft.com/?id=271624

--
Tibor Karaszi, SQL Server MVP
http://www.karaszi.com/sqlserver/default.asp
http://www.solidqualitylearning.com/

Show quoteHide quote

<pshro***@gmail.com> wrote in message news:1116388073.318788.301230@f14g2000cwb.googlegroups.com...
> Thanks for your reply. Would I look at the size of syscacheobjects with
> sp_spaceused to determine the amount of memory that the procedure cache
> is using?
>

18 May 2005 10:30 AM

Kalen Delaney

No, syscacheobjects is a pseudo-table and takes no space on disk, so
sp_spacedused reports 0. The size of the plans is in the pagesused column,
so you can just retrieve the SUM of pagesused.

--
HTH
----------------
Kalen Delaney
SQL Server MVP
www.SolidQualityLearning.com

<pshro***@gmail.com> wrote in message

Show quoteHide quote

news:1116388073.318788.301230@f14g2000cwb.googlegroups.com...
> Thanks for your reply. Would I look at the size of syscacheobjects with
> sp_spaceused to determine the amount of memory that the procedure cache
> is using?
>