I think I have a problem with the execution plan caching in context of prepared statements. Please comment and advise.
When caching a new execution plan SQL Server apparently takes into account the actual query parameters and the current situation of the SQL Server (open transactions, transaction locks, current workload and so on). That can cause the same prepared statement to be executed verry badly with other parameters.
I am having trouble with a production system where some queries more or less suddenly start running extremly bad. The reason is an execution plan which might be optimal for some cases but is in general verry bad. Forcing a recompile of execution plans either by updating statistics or running sp_recompile solves the problem for some time. But after an unpredictable time the bad execution plan is comming back. Probably the good execution plan might also be reinstalled after som time but I cannot wait for this to happen.
The factor between good and bad execution plan is about 160 and increasing (30ms vs. 5000ms).
I am working on tuning the procedure cache hit ratio for my server. We haveadded 4 Gb of memory to the server, which has helped. In addition, I have runthe DBCC FREEPROCACHE, which helped for a couple of days to get the hit ratioup to about 84% (from 68%).When I use the performance monitor on the server and look at SQL Server CacheManager:Buffer Hit Ratio, I see that the Prepared SQL Plan is around 97%, butthe Procedure Plan hit ratio is down around 55%. I've done some research ondifferent tuning techniques, but can't seem to find 1. a clear definition ofthe difference between the prepared sql plan and the procedure plan and 2.other than adding memory and running dbcc freeprocache, how can I get theprocedure plan cache raised? I do know that there are some procedures thatneed to be modified to be called fully qualified (e.g. exec dbo.sp_###instead of exec sp_###), but I don't think that those will increase theprocedure plan by 30% or more.Any insight you can give would be greatly appreciated.Thanks,Michael--Message posted via SQLMonster.comhttp://www.sqlmonster.com/Uwe/Forum...eneral/200511/1
The benefit of the actual execution plan is that you can see the actual number of rows passing through each step - compared to the estimated number of rows.But what about the "cost percentages" ?I believe I've read somewhere that these percentages is still just an estimate and is not based on the real execution.Does anyone know this and preferable have a link to something that documents it?Thanks
I'm new to SQL server but familiar enough with databases to know this doesn't seem right. Here's the situation: I have a table with real estate property information. There are about 650,000 rows in it. I have a nonclustered non-unique index on the city where the property is located. There are about 40 unique values in this index.
I do a simple query like: SELECT city,address from propinfo where city= 'CARLSBAD'. The query will return about 4,000 rows. The problem is that the execution plan that it chooses is to do a full table scan. I.E. Even though there is an index on City, it chooses to look through 650,000 rows rather than do an index seek. Something sounds inefficient here. BTW, this happens in both SQL 7 and SQL 2000. Can anyone explain why this happens? I've got to think that SQL Server is more efficient here.
Hello, I have been looking at the execution plan for a procedure call and the select, compute scalar, stream aggregates, constant scan, nested loops, asserts are all at 0% cost, the PK costs are 2% apart from a rogue 7% and a few 20%, tables scans are all at 23%. The query cost realtive to the batch is 100%. What does this all mean? I have put non-clustered indexes on all the table attributes that are involved in the select statements but this has made no difference, i am guessing this is because my tables are not heavily populated and i may have seen a difference if i had thousands of entries in the tables the select statements acted on, is this assumption correct? Does anyone else bother using the execution plan to tweak there DB or is it a negligible tool?
In sql server 2005 management studio where do I find the option to run the sql query in the query analyser and also show the execution plan? At present I see the option under Query menu which is "Display estimated Execution plan" which only shows the plan but does not execute the query.
Does anyone know of a good way to copy the execution plan when using "Include Actual Execution Plan"? I often need to copy this and mail it.
I know I can use PrintScreen button, but I need a more efficient way to do this. If I just could rightclick the execution plan and select "Copy" and get complete plan it would be great.
Which of the following does NOT cause the execution plan of a query to berecompiled ?- new column is added to a table accessed by a query OR- index used by a query has been dropped from the database OR- query perfoms a join to return data from multiple tables OR- significant amount of data in a table has been mofified
Hi,I have a table-valued user defined function (UDF) my_fnc.The execution of statement "select * from my_fnc" takes much longertime than runnig the code inside my_fnc (with necessary changes).What can be the reason?How can I see an execution plan used for UDF?Thanks a lotMartin
Hi all I'm using Sql server 2000 and sometimes i need to run my Queries in Query analizer before using them in my application just to test them...BUT most of the time when i run a query in query-analizer for second time ,query analizer populates the result (records) more quicker then the first time. Apparently it caches the query !!! i don't know but for some reasons i dont't want this , so how can i prevent Query-analizer from doing so? Thanks in advance. Regards.
I have a report which have multiple parameters. By default all the parameters are selected. Now I am scheduling a Cache with Report Execution Snapshot. Now if I try to view the report, it disables the prompt selection. It means user cant change the selection. Is it the expeceted behaviour of the exectuion snapshot? Thanks, S Suresh
Hi,I want to access the real execution plan via my webapplication after I have executed an SQL statement. I know how to get the estimated execution plan:1 cmd.CommandText = "SET SHOWPLAN_XML ON";2 cmd.ExecuteNonQuery();3 4 cmd.CommandText = myStatement;5 SqlDataReader dataReader = cmd.ExecuteReader();6 7 String plan = String.Empty;8 9 while (dataReader.Read()) {10 plan += dataReader.GetSqlString(0).ToString();11 }12 13 cmd.CommandText = "SET SHOWPLAN_XML OFF";14 cmd.ExecuteNonQuery();I want do compare the estimated costs with the real costs of the same statement. If I change code line 1 an 13 to "SET STATISTICS XML [ON|OFF]" the string "plan" will contain the result of the submitted SELECT statement, but I just need to get the plan and not the result itself. Thanks in Advance,Dominik
I wanted to know whether we have an execution plan enabled in SQL 6.5 as we have it in SQL 7.0 and SQL 2000 . I.e when we execute a query and if we enable ' show execution plan 'then it creates a map and shows the vital statistics . If that is available on SQL 6.5 then i am missing that tool .
How can i have it installed on my SQL 6.5 server ??
I want to know how to analyze query execution plan for complex queries and what information is useful from that for improving the performance. I have gone through details in some sites like www.like sql-performance.com (http://www.sql-server-performance.com/query_execution_plan_analysis.asp), where it was more generic. I want more info regarding this.
Can any one tell about the resources for this or do you have any white papers or documents, which you can share with me.
I am experiencing performance problems with one of my stored procedures. When the stored procedure is first compiled an executed, it behaves as expected (it usually takes 1 or 2 seconds to complete). But its performace it is degradated, so in 1 day, it usually takes 120 seconds to complete !!!. Once the stored procedure is compiled, its performance it is then the expected.
It is a complex stored procedure with two integer parameters with only one select, but composed by multiple views and sub-queries. We have been trying to break the query into small pieces using temporary tables but without success. The SQL Profiler shows an unusual number of reads when it goes wrong (more than a million reads).
I think the problem is in the execution plan. I know than compiling the stored procedure, the problem is fixed, but I do not know exactly when and why it starts to happen.
The stored procedure is running under the following configuration:
- Microsoft SQL Server Standard Edition (64-bit). - Version: 9.00.1399.06 - RAM 16 MB - 8 CPUs
The cost of query with usage of functions is as same as that of withoutfunctionsIn the below code, the query cost of insert is 0.02% and two selectstatements costs same 0.04%Declare @t table(mydate datetime)Declare @i intset @i=1while @i<=5000Begininsert into @t values(getdate())set @i=@i+1EndSelect mydate from @tSelect convert(varchar,mydate,112) from @tBut I thought usage of convert function will take more query costWhat do you think of this?Madhivanan
We've got as slightly unusual scenario happening whereby a statement ispassed to SQL which consists of two parts.BEGIN TRANSACTIONDELETE * FROM WhateverBULK INSERT INTO Whatever...(etc)COMMIT TRANSACTIONThe first is a deletion of the data and the second is the bulk insertof replacement data into that table. The error that we see is aviolation of the primary key (composite).The violation only happens if we run both processes together. If we runone, then the other, it works fine. If we set a line by line insert, itworks fine.My suspicion is that the execution plan that is being run is mostlikely working the two parts in parallel and that the records stillexist at the point that the insert is happening. Truncate is not anoption. The bulk insert was added for performance reasons. There is anoption of trying the bulk insert, and if that fails, do the line byline insert, but it's far from ideal.I think we can probably wrap this into two individual transactionswithin the one statement as follows :BEGIN TRANSACTIONDELETE * FROM WhateverCOMMIT TRANSACTIONBEGIN TRANSACTIONBULK INSERT INTO Whatever...(etc)COMMIT TRANSACTIONWill this give sufficient hint to SQL about the order it processes itso that it completes as we intend and not as it sees being the mostefficient method ?Or, is there a better approach to this ?I've seen that some hints can be passed to SQL for optimizing, but myunderstanding was that it was always better to trust the optimiser andre-work the query as needed.With the server having two processors, is it feasible that one is doingone part and the other processor the other part in parallel ? Willtelling it to use a single processor be worthwhile looking at ? MAXDOP1 ?Finally, I'd imagine that the insert is quicker to process than thedeletion. Is this correct ?ThanksRyan
Using SQL Server 2000 SP4.There is a relatively complex stored procedure that usually completes inless than 20 seconds. Occasionally it times out after 180 seconds. The SPis called via ADO 2.8, using adCmdStoredProc command type. If I useProfiler to capture the EXEC that ADO sends to run the procedure, and runthat from QA, the procedure completes in less than 20 seconds as it should.The procedure is created WITH RECOMPILE. One additional twist is thatsp_setapprole is called from the client before running the procedure inquestion. This may be irrelevant, because even if I include the samesp_setapprole call when running the procedure from QA, it still executesquickly, and even if I comment out the call to sp_setapprole in the clientcode, the proc still times out when run from the client.The only thing that fixes it, at least for a day or two, is DBCCFREEPROCCACHE. So it appears that a bad plan is somehow stuck in memory andis only used when the procedure is called from the client app, and is notflushed even though the procedure was created WITH RECOMPILE.Other than scheduling the DBCC call to run every night, is there anythingelse I could try to get this resolved? Thanks.--(remove a 9 to reply by email)
I was hoping someone could shed some light on wierd situation i'm experiencing. I have the following query:
select count(*) LeadCount from auto_leads al where received > dbo.GetDay(GetDate())
dbo.GetDay simply returns a smalldatetime value of today's date.
Now I recently got thrown into a data mess and for some reason this query takes 8 seconds to run. Now the first thing I did was update the stats on the Received column of this auto_leads table. I re-run the query and I'm still getting 8 seconds. I look at the execution plan I can make figure out why this is happening.
I then change the above query so the filter received > dbo.GetDay(GetDate()) is now just received > '5/31/2006' and the query comes back immediately. This doesn't make sense to me because the GetDay function is really simple and comes back immediately. I then try the following query to confirm it isn't a problem with the GetDay function:
declare @Today DateTime
set @Today = dbo.getday(GetDate())
select count(*) leads from auto_leads al join type_lead_status tls on (tls.type_lead_status_id = al.type_lead_status_id) where received > @Today
Sure enough, the query came back immediately. Next thing to go through my mind is that the query execution plan has been cached by SQL Server using the execution plan from before I updated the stats on the received column. So I executed sp_recompile 'auto_leads' and tryed the original query again. Still taking 8-10 seconds to come back.
So my question, is why when I remove the GetDay function call in my query filter is the query slow, as opposed to me just passing a variable into the query? Thanks!
I'm new to sql server 2005 and was reviewing the execution plan on one of my queries.
I have a query that selects about 62,000 rows from a table of about 20 million
I see there is a index seek indicated but further down the execution plan I see that a large percent is being assigned to a RID LOOKUP on the same table.
Should I be concerned with this and if so, what would you recommend I do to correct it?
We migrated our database from SQL Server 2000 to Yukon last week. Now, when we run our application it has slowed down. We analyzed some stored procedure and they seems to have degarded. The execution plan has changed. Now, this looks like there's lot of work if we have to tune each query w.r.t the new execution plan. Our application has around 4000 stored procs. Is anyone aware of some generic pattern or solution such that these exection plans problem can resolved? Also, does the new execution plan ensure that the once we tune stored procs will perform better than SQL Server 2000.
Need help on this, otherwise it seems we might have to move back to 2000.
I€™m having a test regarding to the image data type. The test program is written with sql native api and just update the image data type column, but I looked the SQL Compilations/sec and Batch Requests/sec counters in SQLServer:QL Statistics using Perfmon, both values are almost the same. It seemed whenever the stored procedure is called, SQLServer compiles it and makes execution plan again. But when I had a test without image data type, SQL Compilation/sec was 0. SQL version is Microsoft SQL Server 2005 - 9.00.3054.00 (Intel X86) (Build 2600: Service Pack 2).
Is SQL server working the way expected or am I missing something?
Hi, I want to know if I can execute a set of batch statements ( basically create statements) during the installation of our product. I have used Access with ADP connected to MS SQL Server 7.0 and I am trying to merge the database and table creations with the installation procedure. Please tell me how I can go ahead with this. (I have tried using the installshield and some other similar products with not of much use). Thanks in advance, Mangala
I want to find out more information about the execution plan. I saw Parallelism (Gather Streams) in the excution plan. In what situation we can see this icon?If in case if we need to avoid how can we avoid this?
We have got a query for fine tuning and it is using lot of CTE ,how can i check the execution plan of that?
CREATE VIEW Mercy AS with ADR as ( SELECT urpx.RoleID , urx.UserID FROM [DBA].dbo.URPX WITH ( NOLOCK ) INNER JOIN [DBA].dbo.URX WITH ( NOLOCK ) ON urpx.RoleID = urx.RoleID WHERE PermissionID = '1' ), SDR as (
-- Collect the roles that a configured with Sales Team Create permission
-- This will include Sales Director , Suite Admin,
SELECT urpx.RoleID FROM [DBA].dbo.URPX WITH ( NOLOCK ) INNER JOIN [DBA].dbo.URX WITH ( NOLOCK ) ON urpx.RoleID = urx.RoleID LEFT OUTER JOIN ADR ON ADR.UserID = urx.UserID WHERE ADR.RoleID IS NULL AND PermissionID='2' )
I have an execution plan that is huge, the pdf it generates if I print it is over 1000 pages. Is there a way to change the graphical plan into a table, so I can sort the %, and find the items that are taking the longest?
if t-sql query is perfectly run in development and when I execute in production at that time I want to use execution plan which is in development . so how I can do using cache? I know about hint we can use hint USE_PLANE. but I want to do with cache .
