Hi!
I was assigned to solve performance problems for an application. I fired up Sql Server profiler and started a trace. Downloaded Sql Server Trace Analyzer. It's a trial version so it's very limited. What I found is that one stored procedure generates almost 400 000 reads everytime it's used and it's used everytime the user wants to see his orders. I've tried to translate the t-sql to english from swedish, it looks something like this:
select top 100
o.orderid,
o.name,
o.latestdeldate,
os.name as OrderStatus,
os.orderstatusID,
p.placeID,
p.name as place,
p.address,
p.city,
a.name as worktype,
noOfActions=(select count(*) from actions a where a.order_orderid=o.orderid),
noOfServiceObjects = (select count(*) from Serviceobject s, Actions a where s.Place_PlaceID = o.Place_PlaceID and a.order_orderid = o.orderid and a.Serviceobject_serviceobjectid = s.serviceobjectid),
...
...
...
It has 8 select count(*) in the select statement then in the where statement it has 2 more select count(*).
I know it's very difficult for you to come up with a solution but do you know a better way than to use select count(*) everywhere? The count is used for to show different status flags on the website.
If I'm doing a dirty reads and a someone updates a record when I'm trying to read it is it possible to read both the old and new records thereby retrieving two records?
How can You find the reads and writes per second of your hard drives in sql. I am reading my SQL book and it says that your average disk should have 125 or less i/o's. And it gave the forumal but as mentioned I don't know how to find the reads and writes.
server: QAT on clustering server ----> 23 seconds ---------------------------------------------------- SS 2000 developer edition SP4 win NT 5.2 (3790) SP4 MeM 7935 MB processors 4 root directory C:program files... use a fixed memeroy size 640 MB
reserve physical memory for sql server minimum query memory 1024 kb
use all available processors minimum query plan threshold for considering 5
PROFILER READS = 5234
server: MILLER ----> 3 seconds ---------------------------------------------------- SS 2000 developer edition no service pack win NT 5.2 (3790) SP4 MeM 2047 MB processors 4 root directory f:MSSQL$INAQAT
dynamically configure sql server memory
use all available processors minimum query plan threshold for considering 5 PROFILER READS = 598
---------------------------------------------------- Making story short. I got an application that hits only 1 database called RECORDS. I'm getting different duration when running an application. 23 and 3 seconds. Same database, same objects and same application. SERVER QAT is our staging server, means lots of databases SERVER MILLER is just a server i just assembled, means just one database (RECORDS).
Not sure if it's because it's a clustering server that is causing the issue nor the reads. If its the reads, what is causing it? Do you think is the how the memory is configured?. Will the experts pls stand up?
So I€™m at a dead-end looking for the reason behind the following behavior. Just to make sure no one misses it, the 'behavior' is the difference in the number of reads between using sp_executesql and not.
The following statements are executed against a SQL 2000 database that contains >1,000,000 records in the act_item table. They are run using Query Analyzer and the Duration and Reads come from SQL Profiler
SQL 2: DECLARE @Priority int DECLARE @Activity_Code char(36)
SET @Priority = 0 SET @Activity_Code = '46DF335F-68F7-493F-B55E-5F9BC6CEBC69' update act_item set Priority = @Priority where activity_code = @activity_code
Reads: ~160 Duration: 0 ms
Random information:
Activity_code is an indexed field on the table, although it is not the primary key. There are a total of four indexes on the table, none of which include the priority as one of the fields. There are two triggers on the table, neither of which is executed for this SQL statement (there is an IF UPDATE(fieldname) surrounding the code in the trigger) There are no foreign relationships I checked (using perfmon) to see if a compilation/recompilation was happening. No it's not. Any suggestions as to avenues that could be examined would be appreciated.
Hello, im using sqldatareader to read my data and whenever time i loop through the reader it starts from second row why is that? here is my code:while (reader.Read()){hinfo.Name = reader["_name"].ToString();hi.Add(hinfo);} i look at the database and i have two rows but its reading only the second row, skiping the first row
I have a set of triggers that log the history of changes to a table - i.e. I record inserts, updates, deletes (pretty standard audit stuff I suppose). I want to also log reads on that data. If I were using sprocs for reading data, this would be relatively painless, but I am using an O/R mapper to handle my data access, which writes dynamic sql at runtime (and I don't want to use sprocs with it) and then sends it down to the DB. Is there a way I can intercept reads and log them to the same table I am logging other actions? I know very little about the new capabilities of SQL Server 2005, but I would think I could somehow, maybe via the new CLR capabilities or similar, get access to these types of events within the database? Anyone? I know I could always do this higher up in the application layers, but I would like to keep all of this at the database level if possible....Thanks,
Is there a way to get a total count of all SELECT, UPDATE, DELETE and INSERT statements to a SQL Server 6.5 database during a 12 hour period? I'm thinking maybe someone knows of a software that reads the log or monitors the server... I've been looking at the performance monitor and, although it has good information, it doesn't capture DML's.
I'm trying to insert all the rows from a table to a new table. (insert A select * from AA) The reads on Profiler shows ar really high value (10253548).
First I created a unique clustered index and the reads shows (3258445), then I created a non clustered index expecting to have lower reads. Instead the reads shows (10253548).
I read creating indexes helps reduce reads. But it's not happening. Any ideas what is going on?
Can any of can explain, what the "Reads" column in Profiler exactly mean ? I'm not comfortable with the explanation given in BOL.
"The number of read operations on the logical disk that are performed by the server on behalf of the event. These read operations include all reads from tables and buffers during the statement's execution"
For the same procedure with same parameters, if the server is not loaded much, the Reads are in a few hundreds, but when there are more than 1000 concurrent users, why it is going to millions ? What other parameters affecting this reads ? And how can I reduce it ?
Environment: SQL Server 2005 64-bit Enterprise Edition on Windows Server 2003 R2 Server x64 Enterprise Edition SP2
I have been seeing a basic scenario of a write transaction appearing to unexpectedly lock-out reading.
The database has isolation set to "READ COMMITTED".
The scenario is:
1.) Start a transaction (for doing a write)
2.) Do a read before the transaction (for doing the write) is committed (e.g. sqlCommand2.ExecuteReader()).
--> the code will appear to lock-up (then time out).
I see the same behavior if I step through the "write" code with the debugger (to a point after the transaction is started, but before it is committed), and run a "SELECT * FROM" type query from Microsoft SqlServer Management Studio.
Following is the code sample demonstates the issue.
Thoughts on how to resolve the issue (to let me do "read committed" reading of the database table)?
Thanks!
Andy
Module Transaction
Sub Main()
Dim exception1 As Exception
Try
' Create/Open Database Connection
Dim sqlConnection1 As New System.Data.SqlClient.SqlConnection("Server=GRB-AB;Database=Transaction;Trusted_Connection=True;")
sqlConnection1.Open()
' Start transaction
Dim sqlTransaction1 As System.Data.SqlClient.SqlTransaction = sqlConnection1.BeginTransaction()
' Set Parent record
Dim sqlCommand1 As New System.Data.SqlClient.SqlCommand("INSERT INTO Parent (Name) VALUES ('ParentValue');", sqlConnection1)
sqlCommand1.Transaction = sqlTransaction1
sqlCommand1.ExecuteNonQuery()
' Get Id from parent record (note: this code assumes the table was empty when this program starts)
sqlCommand1 = New System.Data.SqlClient.SqlCommand("SELECT Id FROM Parent;", sqlConnection1)
sqlCommand1.Transaction = sqlTransaction1
Dim parentId As Integer = CType(sqlCommand1.ExecuteScalar(), Integer)
'
' Do reading test to test concurrently reading table being written to
'
' Create/Open Database Connection for reading test
Dim sqlConnection2 As New System.Data.SqlClient.SqlConnection("Server=GRB-AB;Database=Transaction;Trusted_Connection=True;")
sqlConnection2.Open()
Dim sqlCommand2 As New System.Data.SqlClient.SqlCommand("SELECT Id FROM Parent;", sqlConnection2)
sqlCommand2.ExecuteReader()
Dim i As Integer
While (sqlCommand2.ExecuteReader.Read = True) ' <===== LOCKS UP HERE **************
i = i + 1
End While
'
' End reading test
'
' Set child record
sqlCommand1 = New System.Data.SqlClient.SqlCommand( _
I have written a same stored proc in TSQL and SQL CLR which basically takes an input xml and returns xml document. In SQL Profiler, I am getting reads value about five times more for the CLR. Does anyone has any idea why the CLR is doing more reads than TSQL? Thanks in advance.
Just migrated application from Oracle to SQL and we are seeing alot of deadlocking and blocking. I did notice that app seems to be passing isolation level of repeatable read. Attached is a .doc of one of the deadlocks, is there a way to avoid these in the repeatable read isolation level? This example is a select with two tables, using NCI's that cover the where, and a insert doing just a clustered index insert. Is this simply try to get rid of the repeateable read if not needed, guess have to check with vendor on that or is there a way to get this to not deadlock using repeatable read?
I'm trying to figure out why my sqlserver is flatlined on the CPU. I'm doing a trace and can't help but notice this, with crazy high reads. I'm not sure what this is? It doesnt look good to me, altho maybe its nothing. Any info is much appreciated.
I am running a profiler trace against a database and noticed that thereads column always shows 0. When running the same trace againstanother machine I get back values in the reads column. I took a querythat profiler reported as having 0 reads and ran in in query analyzerwtih STATISTICS IO on and confirmed that there are in fact reads:Table 'tt_cawardalloc'. Scan count 1, logical reads 8, physical reads0, read-ahead reads 1.Table 'tt_clineitem'. Scan count 10, logical reads 125208, physicalreads 1540, read-ahead reads 2995.Table 'tt_contractitem'. Scan count 32, logical reads 676, physicalreads 0, read-ahead reads 0.Table 'tt_contract2'. Scan count 3, logical reads 121, physical reads4, read-ahead reads 0.I am on SQL 2000 sp3a. Any help appreciated.Thanks!
I'm running the same query on two different PCs and tracing results in Profiler on my PC. When executing the query on PC1 - the total number of reads is 200000. When executing the same query on PC2 - the toal number of reads is 13000. It is almost 15 times more reads when executing query on PC2. The executed query is same on PC1 and PC2. Any reason for this?
I'm trying to analyse that query and reduce the number of logical reads as it's is too high but then I get completly different result on different PC.
Lets say user A accesses a record and is making an update to a column... next user B accesses the same record and makes an update to the same column and saves the data... how can user A check to see if an update has been made to prevent overwriting the data..
Is there a query statement that user A can write to check for this?
I understand locking can be used to prevent this but is there an alternative to locking.
Why is there often such a dramatic discrepancy between the logical reads recorded in the trace file versus the output of STATISTICS IO?
In the server-side trace I have running I found a reporting procedure that shows having 136,949,501 reads (yes, in hundreds of millions), and it's taking 13,508 seconds to complete.
So I pull the code from the trace and execute it via SSMS - it runs < 1 second, and only generates about 4,000 reads (using various different parameters I get the same result)
All my queries are being blocked while the tables are being replicatedand it is causing some 2 minute blocking. Is there a way for theReplication to allow dirty reads because I really don't care aboutthat, I would rather have dirty reads than 2 minute waits.Thanks.
How do I determine which method I should use ifI want to optimize the performance of a database.I took Northwind's database to run my example.My query is I want to retrieve the Employees' Firstand Last Names that sold between $100,000 and$200,000.First let me create a function that takes the EmployeeIDas the input parameter and returns the Employee'sFirst and Last name:CREATE FUNCTION dbo.GetEmployeeName(@EmployeeID INT)RETURNS VARCHAR(100)ASBEGINDECLARE @NAME VARCHAR(100)SELECT @NAME = FirstName + ' ' + LastNameFROM EmployeesWHERE EmployeeID = @EmployeeIDRETURN ISNULL(@NAME, '')ENDMy first method to run this:SELECT EmployeeID, dbo.GetEmployeeName(EmployeeID) ASEmployee, SUM(UnitPrice * Quantity) AS AmountFROM OrdersJOIN [Order Details] ON Orders.OrderID =[Order Details].OrderIDGROUP BY EmployeeID,dbo.GetEmployeeName(EmployeeID)HAVING SUM(UnitPrice * Quantity) BETWEEN100000 AND 200000It's running in 4 seconds time. And here are theStatistics IO and Time results:SQL Server Execution Times:CPU time = 0 ms, elapsed time = 0 ms.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 0 ms.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 0 ms.SQL Server parse and compile time:CPU time = 17 ms, elapsed time = 17 ms.(3 row(s) affected)Table 'Order Details'. Scan count 1, logical reads 10,physical reads 0, read-ahead reads 0.Table 'Orders'. Scan count 1, logical reads 21,physical reads 0, read-ahead reads 0.SQL Server Execution Times:CPU time = 3844 ms, elapsed time = 3934 ms.SQL Server Execution Times:CPU time = 3844 ms, elapsed time = 3935 ms.SQL Server Execution Times:CPU time = 3844 ms, elapsed time = 3935 ms.SQL Server parse and compile time:CPU time = 0 ms, elapsed time = 0 ms.Now my 2nd method:IF (SELECT OBJECT_ID('tempdb..#temp_Orders')) IS NOT NULLDROP TABLE #temp_OrdersGOSELECT EmployeeID, SUM(UnitPrice * Quantity) AS AmountINTO #temp_OrdersFROM OrdersJOIN [Order Details] ON Orders.OrderID =[Order Details].OrderIDGROUP BY EmployeeIDHAVING SUM(UnitPrice * Quantity) BETWEEN100000 AND 200000GOSELECT EmployeeID, dbo.GetEmployeeName(EmployeeID),AmountFROM #temp_OrdersGOIt's running in 0 seconds time. And here are the Statistics IOand Time results:SQL Server Execution Times:CPU time = 0 ms, elapsed time = 0 ms.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 0 ms.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 0 ms.SQL Server parse and compile time:CPU time = 0 ms, elapsed time = 0 ms.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 0 ms.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 0 ms.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 0 ms.SQL Server parse and compile time:CPU time = 0 ms, elapsed time = 0 ms.Table '#temp_Orders0000000000F1'. Scan count 0, logicalreads 1, physical reads 0, read-ahead reads 0.Table 'Order Details'. Scan count 830, logical reads 1672,physical reads 0, read-ahead reads 0.Table 'Orders'. Scan count 1, logical reads 3, physical reads 0,read-ahead reads 0.QL Server Execution Times:CPU time = 15 ms, elapsed time = 19 ms.(3 row(s) affected)SQL Server Execution Times:CPU time = 15 ms, elapsed time = 19 ms.SQL Server Execution Times:CPU time = 15 ms, elapsed time = 20 ms.SQL Server parse and compile time:CPU time = 0 ms, elapsed time = 1 ms.(3 row(s) affected)Table '#temp_Orders0000000000F1'. Scan count 1,logical reads 2, physical reads 0, read-ahead reads 0.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 3 ms.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 3 ms.SQL Server Execution Times:CPU time = 0 ms, elapsed time = 3 ms.SQL Server parse and compile time:CPU time = 0 ms, elapsed time = 0 ms.By the way why "SQL Server Execution Times"exists 3 times and not just one time?Summary:The first code is clean, 1 single SELECT statement buttakes 4 long seconds to execute. The logical reads arevery few compared to the second method.The second code is less clean and uses a temp table buttakes 0 second to execute. The logical reads are waytoo high compared to the first method.What am I supposed to conclude in this example?Which method should I use over the other and why?Are both methods good depending on which I prefer?If I can wait four seconds, it's better to reduce the logicalreads in order to provide less Blocking on the live tablesin a heavily accessed database?Which method should I choose on my own database?Calling a function like dbo.GetEmployeeName getsprocessed per each returned row, correct? That meansIf i had a scenario where 1000 records were to be returnedwould it be better to dump 1000 records to a temp tablevariable and then call a function to process each recordone at a time?Or would the direct approach without usinga temp table cause slower processing and moreblocking/deadlocks because I am calling the functionper each row as I am accessing directly from the tables?Thank you
Hi. Periodically I need to run a delete statement that deletes old data. The problem is that this can timeout using ODBC (via the CDatabase and CRecordSet classes in legacy code). Also, while its running the delete, the table its operating on is locked and my application can't continue to run and operate on rows not affected by the delete.
Are there any workarounds for this? Can the timeout be set in the connect string?
A table in one of my databases is running very slowly. The IO is very high and below is a printout from the SET STATISTICS IO ON command run on a common query used on the table:
I have a clustered unique index and a nonclustered index on the table. I have ran SQL Profiler and opened the trace in Database Tuning Advisor, DTA displays 0% improvement suggestions. I have a number of statistics on the table and index which are all up to date and fragmentation is less than 1%. I've tried a number of variations on indexes to improve performance but to no avail. There is only one query which runs on the table, and the nonclustered index created on the table did significantly improve performance, however the query still runs at around 23 seconds. The query does bring back a large amount of data however i'm sure there is a way to bring down the IO and logical reads to improve performance.
-- =================== Nonclustered Index ===========================
CREATE NONCLUSTERED INDEX [dta_ix_WebProxyLog_Kaction_clientusername_logtime_uri_mimetype_webproxylogid] ON [dbo].[WebProxyLog] ( [Action] ASC ) INCLUDE ( [ClientUserName], [logTime], [uri], [mimetype], [WebProxyLogid]) WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
-- =================== Query which is called regularly on the table ===========================
SELECT [User] = CASE WHEN LEFT(clientusername,3) = domain' THEN RIGHT(clientusername,LEN(clientusername) - 3) ELSE clientusername END, logtime AS [Date], desthost AS [Site], uri AS [Actual Site] FROM webproxylog WHERE CONVERT(Datetime,CONVERT(VarChar(25),logtime,106),106) BETWEEN '20 apr 2008' AND '14 may 2008' AND(RIGHT(uri,4) NOT IN('.css','.jpg','.gif','.png','.bmp','.vbs')) AND (RIGHT(uri,3) NOT IN('.js')) AND LEFT(mimetype,6) = 'text/h' AND (uri NOT LIKE '%sometext.local%') AND (uri NOT LIKE '%sometext.co.uk%') AND [action] = 9 AND (clientusername IN ('USERNAME')) ORDER BY logtime ASC;
I'm using an application that produce 48000000 reads for one stored procedure and 170 seconds to complete. The same procedure when executed in SQL Analyzer takes only one seconds and 10000 reads.
What is happening here? Where should I look to solve this problem?
I'm doing som performance research, I have a index with following priority: ClientId, Active, ProductId. Active is a bit field telling whether the Product is active or not, it can be inactive products than active, but always at least one active product. When I'm executing
SELECT * FROM [table] WHERE ClientId = [id] AND ProductId IN (1,2,3,5,7,9,20)
I'm getting following result: Scan count 1, logical reads 490
When I'm leading SQL Server to the right paths by including the to possible values in Active by executing the following SQL:
SELECT * FROM [table] WHERE ClientId = [id] AND ProductId IN (1,2,3,5,7,9,20) AND Active IN (0,1)
I'm getting following results: Scan count 14, logical reads 123
With this information, which version would you say is fastest and why?
When I was running this query 1000 times with different ClientId I got a average time of 172 ms for the first query, and 155 ms for the second one. I have been told that scan count is very expensive... out of this example it seems that the cost of 1 scan count is like 20 logical reads?
I recently installed an evaluation copy of SQL Server 2000 on a machine with Windows Server 2000. Nearly all actions in the enterprise manager cause 15-second to 60-second loud, monotonous reads of the hard drive. Anything from expanding a tree item to creating a table.
This makes the product more or less unusable. What could be causing it?
All other software on the machine works normally, and nothing else seems wrong.
We are in the process of moving existing clustered SQL server databases to AWS. There is one major database that has intensive reads and writes transactions. I'm wondering what is the best design to optimize the performance for both R/W since we have constant issues historically with the current environment when massive updates are happening. Reads shall have higher priority over writes.
I have inherited a database that is over-indexed, i.e. there are sometimes 10-20 indexes on a table. The performance is at times not great due to blocking from long running queries. I want to clean up the indexes as a starting point.
Through a query I found some time ago on the SQLCat blog I have discovered a large number of indexes in the database that have a huge disparity between reads and writes. The range of difference is sometimes almost 2 million more writes than reads. Should I just drop the indexes that have say, more than 100,000 more writes than reads and then see what the Missing Index DMVs tell me after a few days of running without those indexes?
In some cases there are a few hundred thousand reads but maybe a million writes on the index. Thus, there are a fair number of reads happening, just not in comparison to the number of writes. In some cases there are almost no reads and a million or more writes. I am obviously dropping those indexes. I just am not sure what to do about the indexes that do have a fair number of reads.
