Re: SQl Data Provider Performance Issues

Greg,

Can you provide a complete sample app (with a database file and code)
which would outline the problem. Without seeing any of the code, it's
impossible to tell at this point.


--
- Nicholas Paldino [.NET/C# MVP]
- mvp@xxxxxxxxxxxxxxxxxxxxxxxxxxx

"Greg" <gmichaels@xxxxxxxxxxxxx> wrote in message
news:1186499523.874600.117200@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
On Aug 7, 9:05 am, Greg <gmicha...@xxxxxxxxxxxxx> wrote:
On Aug 7, 7:39 am, "Nicholas Paldino [.NET/C# MVP]"





<m...@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
Based on your other posts, it seems like it's not so much the
individual
fills, but rather, the fact that you have so many of them. Are the
numbers
you gave for one individual call to Fill the data set, or is it for all
of
them?

I believe you are on the right track when you say that it is the
interaction of the app with the database, but you haven't given many
details
about that. I don't think it is the SqlDataAdapter in itself. Have
you
tried just executing the command, getting a reader, and then cycling
through
it (columns and rows) to see what the performance there is? Or, maybe
just
filling one dataset, and measuring the performance there?

--
- Nicholas Paldino [.NET/C# MVP]
- m...@xxxxxxxxxxxxxxxxxxxxxxxxxxx

"Greg" <gmicha...@xxxxxxxxxxxxx> wrote in message

news:1186468407.360935.9580@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

On Aug 6, 7:36 pm, "Nicholas Paldino [.NET/C# MVP]"
<m...@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
Greg,

It seems like something else is definitely going on here. I
can't
see
the code you are using, so I can't tell what is going on on the .NET
side.
However, from what you have shown of the execution tree, it seems
like
you
haven't optimized the ValidValues_HLAHighRes table for this query.
It's
doing an index scan, which is not really what you want. You should
probably
index it so that an index seek will be performed. If you have a
large
number of records in the ValidValues_HLAHighRes table, then that
could be
impacting performance.

Of course, there is some overhead in populating a data set, but
from
what you are mentioning, it doesn't seem like it should be that
much.
Assuming a reasonable number of columns in the table, a few hundred
rows
really shouldn't take that long.

--
- Nicholas Paldino [.NET/C# MVP]
- m...@xxxxxxxxxxxxxxxxxxxxxxxxxxx

"Greg" <gmicha...@xxxxxxxxxxxxx> wrote in message

news:1186446086.255319.90070@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

I am trying to fill strongly typed datasets with data from a
SQLServer
DB. The data is used as a datsource for drop down lists. The data
adapters are configured to perform only selects, no inserts,
updates,
or deletes. When I call the Fill method to load the dataset, CPU
Usage
goes to 100%. These simple queries take forever to complete. We
are
only dealing with a few hundred rows. My windows form takes
minutes to
load because of the poor performance. The select statements
require 2
inner joins to aquire the required data columns. Similar fills
using
selects with a single inner join perforn OK. I tried running the
SQl
Profiler to analyze the problem. When I run the query from SQL
Query
Analyzer, SQL Profiler shows a duration of 30. When run from
withing
the C# .net application duration jumps to 9000 or 300 times
slower.
Reads stay the same. SQL execution plan is the same when using
Query
analyzer versus the application. It appears the SQL Data Provider
is
the source of the poor performance. Here is the SQL Select that
performs poorly:

SELECT H.HLAHighResID, H.HighResName, H.HLALowResID, L.LowResName
FROM ValidValues_HLAHighRes H
INNER JOIN ValidValues_HLACategories C ON H.HLACategoryID =
C.HLACategoryID
INNER JOIN ValidValues_HLALowRes L ON H.HLALowResID =
L.HLALowResID
WHERE (C.CategoryName = 'A')

Execution Tree
--------------
Nested Loops(Inner Join, OUTER REFERENCES:([H].[HLALowResID]))
|--Nested Loops(Inner Join, WHERE:([H].[HLACategoryID]=[C].
[HLACategoryID]))
| |--Index
Seek(OBJECT:([test].[dbo].[ValidValues_HLACategories].
[IX_ValidValues_HLACategories_1] AS [C]), SEEK:([C].
[CategoryName]='A') ORDERED FORWARD)
| |--Clustered Index Scan(OBJECT:([test].[dbo].
[ValidValues_HLAHighRes].[PK_ValidValues_HLAHighRes] AS [H]))
|--Clustered Index
Seek(OBJECT:([test].[dbo].[ValidValues_HLALowRes].
[PK_ValidValues_HLALowRes] AS [L]), SEEK:([L].[HLALowResID]=[H].
[HLALowResID]) ORDERED FORWARD)

Here is the SQL Select that performs OK even though duration
doubles:

SELECT L.HLALowResID, L.LowResName
FROM ValidValues_HLALowRes L
INNER JOIN ValidValues_HLACategories C ON L.HLACategoryID =
C.HLACategoryID
WHERE (C.CategoryName = 'A') ORDER BY L.LowResName

Execution Tree
--------------
Sort(ORDER BY:([L].[LowResName] ASC))
|--Nested Loops(Inner Join, WHERE:([C].[HLACategoryID]=[L].
[HLACategoryID]))
|--Index
Seek(OBJECT:([test].[dbo].[ValidValues_HLACategories].
[IX_ValidValues_HLACategories_1] AS [C]), SEEK:([C].
[CategoryName]='A') ORDERED FORWARD)
|--Clustered Index Scan(OBJECT:([test].[dbo].
[ValidValues_HLALowRes].[PK_ValidValues_HLALowRes] AS [L]))

Thes are quite similar excpet for an extra INNER JOIN. What can
the
souce of this poor perfomance be? Why does CPU get pegged at 100%
when
I run these selects from the application? I thought trhe SQL Data
Provider was the preferred choice when accessing a SQL Server DB.
Should I try an OLE Data Adapter?- Hide quoted text -

- Show quoted text -

The HLAHighRes table has 3600 rows, HLALowRes has 140 rows, And
HLACategories has just 6 rows.
The queries run very fast from query analyzer but get terribly slow
when called by the data adapter fill method.
CPU goes to 100%. In my DB experience it performs like it is doing
full table scans with thousands of rows.
All the tables are keyed by an identity column hence the ID in the
column name and have a primary clustered key index.
I tried adding an index on the combo of H.HLAHighResID,
H.HighResName,
H.HLALowResID
which was recommended by the profiler index analyzer. It really did
not effect performance.
Seems to me there is some issue with the interaction of the app and
DB.- Hide quoted text -

- Show quoted text -

40 of the datasets get filled when the form loads. The 1st 10 that use
1 INNER JOIN load OK.
As soon as it hits the next 10 that have 2 INNER Joins performance
goes bad. Then the next 10 datasets have
the 1 INNER JOIN again and performance picks up. The last 10 datasets
filled are back to 2 INNER JOINS and performance goes bad again.
I have a Profiler Trace showing this. I tried changing the order by
putting the poor performers 1st. It did not matter.
The performance was still going from poor to good depending on the
query and not the order.- Hide quoted text -

- Show quoted text -

The numbers I gave were for 1 dataset fill and not all of them. On a
fill by fill basis the performance degrades by a factor of 30 when
running the query in SQL Query Analyzer versus a fill from within the
app. Multiply the degradation in performance by 20 and you can see why
I am frustrated. The app takes forever (minutes) to load.



.

Relevant Pages

  • Re: SQl Data Provider Performance Issues
    ... selects with a single inner join perforn OK. ... When I run the query from SQL ... Analyzer, SQL Profiler shows a duration of 30. ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: Ranking query
    ... I expect that using a named query (as opposed to its SQL) will be OK. ... INNER JOIN qryRepairs AS I2 ...
    (microsoft.public.access.queries)
  • Re: SQl Data Provider Performance Issues
    ... selects with a single inner join perforn OK. ... When I run the query from SQL Query ... Analyzer, SQL Profiler shows a duration of 30. ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: SQl Data Provider Performance Issues
    ... selects with a single inner join perforn OK. ... When I run the query from SQL ... Analyzer, SQL Profiler shows a duration of 30. ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: SQl Data Provider Performance Issues
    ... Can you provide a complete sample app ... selects with a single inner join perforn OK. ... When I run the query from SQL ...
    (microsoft.public.dotnet.languages.csharp)