AssafST
asked on
performance - sql server select vs. select on in memory datatable
Hi,
I’m looking for the best way to do the following thing in terms of performance time/memory usage.
I’ve a windows app. build with C#. the machine is win server 2003 with sql server 2005 DB.
I’ve a DB with a few tables, some have 1000's rows and some 10,000's rows.
in my application I need to do many select commands.
there're 2 approaches I’m aware of and I’m looking for the best one in terms of performance:
1. load the tables into datatables at the beginning of the app. run using SqlDataAdapters and then use datatable.select(xxx) to find the rows I need (no primary key selects usually select on text).
in this method I approach sql server only in the beginning of the app., loads the data to the memory and query on the datatables.
2. each time I’ve to query use a stored procedure on the DB. which means 10000's of queries on the DB.
from test I’ve made the 2nd method looks faster, but on task manager it looks like the sqlservr.exe is working very hard and takes a lot of CPU.
so what is the best way? is there another solution which is better than my solutions?
what about LINQ?
Thanks,
Assaf.
I’m looking for the best way to do the following thing in terms of performance time/memory usage.
I’ve a windows app. build with C#. the machine is win server 2003 with sql server 2005 DB.
I’ve a DB with a few tables, some have 1000's rows and some 10,000's rows.
in my application I need to do many select commands.
there're 2 approaches I’m aware of and I’m looking for the best one in terms of performance:
1. load the tables into datatables at the beginning of the app. run using SqlDataAdapters and then use datatable.select(xxx) to find the rows I need (no primary key selects usually select on text).
in this method I approach sql server only in the beginning of the app., loads the data to the memory and query on the datatables.
2. each time I’ve to query use a stored procedure on the DB. which means 10000's of queries on the DB.
from test I’ve made the 2nd method looks faster, but on task manager it looks like the sqlservr.exe is working very hard and takes a lot of CPU.
so what is the best way? is there another solution which is better than my solutions?
what about LINQ?
Thanks,
Assaf.
ASKER
1. the data is changed/updated at the end of the App. run - this App. is used to update the data in the DB, so there's no fear that the data isn't correct.
2. i need all the tables data in order to run the App., so there're 6 tables with ~10,000 rows each.
2. i need all the tables data in order to run the App., so there're 6 tables with ~10,000 rows each.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
no. i need all the data in order to update 3 tables (not all the records).
i don't do the update in memory but create a long string of SP commands and at the end of the run update the DB.
something like:
string updateDBstr = "exec SP1 p1, p2,p3; exec SP1 p5, p6,p7; exec SP2 p1, p2,p3;" etc...
objCommO.CommandText = updateDBstr;
objCommO.ExecuteNonQuery()
i don't do the update in memory but create a long string of SP commands and at the end of the run update the DB.
something like:
string updateDBstr = "exec SP1 p1, p2,p3; exec SP1 p5, p6,p7; exec SP2 p1, p2,p3;" etc...
objCommO.CommandText = updateDBstr;
objCommO.ExecuteNonQuery()
If you have enough memory on the SQL box you should not worry about that as SQL will put those tables in cache and keep them as long as they are needed (especially on not so large tables as yours). Until SQL 2005 you could have use DBCC PINTABLE sql command to force SQL keep those tables in memory but was discontinued.
you should check your page life expectancy/SQL Server Buffer Cache Hit Ratio and if less than 90% you may need better indexing/stats or more RAM on the box
you should check your page life expectancy/SQL Server Buffer Cache Hit Ratio and if less than 90% you may need better indexing/stats or more RAM on the box
ASKER
1. what do you mean by "put those tables in cache"? put them in datatables as i do?
2. "you should check your page life expectancy/SQL Server Buffer Cache Hit Ratio"
how do i check this?
10x
2. "you should check your page life expectancy/SQL Server Buffer Cache Hit Ratio"
how do i check this?
10x
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
not a final solution
1. Loading the data at startup - what if data has changed ?
2. How big is the return ? Do you return 10 rows, 100 rows, 10000 rows, ... The higher you go, the inneffecient your design is made. Perhaps you can use paging to return the data in smaller pieces ...