Using one instance of SqlConnection in an ASP.NET page

B

Bob

In our new .NET web applications, we try to limit the use of SqlConnection
to just one instance per page, even if there are multiple accesses to
various queries. The thinking behind is that this reduces the need to
getting and returning connections to the pool repeatedly if a page has
multiple calls to the DB, and each one manages its own connection. However,
this does requires more deliberate coding, like calling the
SqlConnection.Close() method in the page's Dispose() method so it's
garanteed that the connection is closed when the page processing is done,
and also is not closed too early. What I'm thinking is whether this is
actually necessary because passing the Connection object into child controls
becomes a pretty big hassle when there are several user controls or custom
controls on the page. If the connection pool management is very efficient
then opening and closing connections repeatly in the code (which is really
getting and returning connections to the pool) wouldn't be a big deal, and
keeping the code simple would be more important. Could anyone give some
suggestions?
 
A

Alex Papadimoulis

Bob, IIRC, so long as you open the connection with the same connection
string, it'll be pooled.

Alex Papadimoulis
 
B

Bob

Alex, thanks for your reply. I understand that the same connection string
reults in connections in the same pool. My question is whether getting and
returning connections to the pool have become so efficient in ADP.NET and
Sql Data Provider that there is no need to put in extra code in order to
reduce the number of getting and returning from say, 3 to 1.

Thanks
Bob
 
B

Brad Williams

Bob said:
In our new .NET web applications, we try to limit the use of SqlConnection
to just one instance per page, even if there are multiple accesses to
various queries. The thinking behind is that this reduces the need to
getting and returning connections to the pool repeatedly if a page has
multiple calls to the DB

Someone can correct me if I'm wrong, but I believe getting and returning
connections from managed connection pool is very efficient -- otherwise it
wouldn't be much of a pooling/caching mechanism!

Brad Williams
 
A

Alex Papadimoulis

Bob,

I don't believe there would be any noticible difference. You may be giving
up a few cycles to do that, but it really is a better practice. Write the
simpler, easier to manage code now. If you see a performance problem, then
optimize.

-- Alex Papadimoulis
 
B

Bob

Hmm, that's still not what I was asking. I understand how and where to
properly close a connection, why it should be done, where to set the pool
size and timeout etc. What I'm trying to get at is that, if I have a page
that needs to make a series of calls to the database, say, call SP 1 to get
some data to populate a drop down, then call SP 2 to get some other data to
populate a datagrid, in the same page a user control on the page calls SP 3
to get some status data to display on the sidebar. So I face a design
decision here. One way is to create one instance of the SqlConnection
object (could be handled by a middle tier but for discussion purpose let's
say I'm doing this directly from the aspx.cs class) and use it throughout
the page, and put the .Close() in the page Dispose() method so the
connection is closed at the end and only at the end of the page processing.
Alternatively, I can design it to let each call create its own instance of
the SqlConnection object and close it immediate after the call is done.
This way, I can wrap each call in its self contained function, and
particularly for the user control on the page, I don't have to get the
connection instance from the main page. This makes the code simpler and the
logic a lot easier to follow. However, the second option would use 3
different connections. Assuming the connections are already in the pool so
it doesn't have to create brand new ones, it would be getting and returning
connections to the pool 3 times, rather than 1 in the first option. So back
to my original questions, if getting and returning connections to the pool
has become so efficient that doing it 2 times more is no big deal at all
then I probably should go with option 2, otherwise I should still do option
1, which has more complicated coding particularly if a middle tier is
involved.

Thanks


Bin Song said:
Hi, Bob

I understand your question now.
The returning of Connection to pool depend on when you close the
connection or the connection time out. The pool size and timeout can be
defined in the connection string.
It is recommended that you always close the Connection when you are
finished using it in order for the connection to be returned to the pool.
This can be done using either the Close or Dispose methods of the Connection
object. Connections that are not explicitly closed might not be added or
returned to the pool. For example, a connection that has gone out of scope
but that has not been explicitly closed will only be returned to the
connection pool if the maximum pool size has been reached and the connection
is still valid.
 
J

Joe Fallon

Bob,
You should use a single connection and just open and close as needed.

The connection in the pool is not really closed. It is just sitting there
waiting for you to ask for it again.
So you lose nothing to "close" a connection and the "open" it. It is already
open.

In my Business Objects I open a datareader to populate the BO and then close
it and re-open a 2nd dr to populate any contained BOs or collections. The
dr.close also "closes" the connection but then I re-open it and poulate the
next object.

HTH
 
B

Bob

Thanks a lot guys. sounds like you all are saying the 2nd option is better,
that is, open and close immediately as needed, even if it means doing it
several times in one request. I also read some other articles which lead to
similar conclusions.

Thanks again for the help.

Bob
 
P

Pablo Castro [MS]

Pooling tends to be very efficient, particularly if you're not using
DTC/COM+ transactions. Integrated security also has an small impact, but is
has been greatly optimized in .NET 1.1.

That said, if you want to squeeze every last bit of performance, keeping the
connection open for the execution of the entire page is certainly better.
While pooling is efficient, there is quite a bit of code that needs to run
in order to grab a connection, and we can potentially hit a lock because of
other threads accessing the pooler.

Regarding the scalability implications of this: is most cases a page is
processed very quickly and "all together" -meaning that there are no long
pauses during the page rendering process. Given that assumption, there is no
benefit on opening/closing the connection at every use within the same page.

--
Pablo Castro
Program Manager - ADO.NET Team
Microsoft Corp.

This posting is provided "AS IS" with no warranties, and confers no rights.
 
R

Rick Spiewak

As a third option, consider using the Microsoft Data Access Application
Block, and let it worry about the rest <g>!
 
J

Jason Coyne Gaijin42

The second option is much better. You should open up a connection for
each function and close it as soon as possible. You will get the same
connection back from the pool 99.9% of the time. More importantly,
OTHER sessions will most likely get that SAME connection, meaning your
database is only seeing one connection. This is infinetly more
scalable.


Regardless of scalability, having one connection and just keeping it
open the whole time is not technically possible for any but trivial
applications, as you can only have one recordset open at a time per
connection. This means that while you are looping through a
datareader, you cannot run any other queries or get any other
datareaders using that connection. This seems like a big step
backwards from classic ADO, but in reality it is not. ADO had the
same limitation, it just opened up a second connection for you
automatically under the covers. The same total number of connections
open is the same, but in ado.net you are aware that you are using two
connections, so perhaps you can redisign code to work more
efficiently.

In my applicaitons I have a connect() function that creates and opens
then returns a connection. I call that at the beginning of each of my
data functions. Then if I am using a dataset, or doing an
executenonquery, I close right inside that funciton. If I am returning
a datareader outside of the function, I always use
CommandBehaviour.CloseConnection and pass it out that way to minimize
impact on my connection pool.
 
A

Angel Saenz-Badillos[MS]

Bob,
A confusing thread, but the message is loud and clear (and completely
correct) Open the connection as late as possible, close it as soon as you
can and in a Finally block (NOT on the page dispose method). The try finally
or "using" construct is the only real way to guarantee your connection will
be closed if an exception happens while your connection is open.
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

No members online now.

Forum statistics

Threads
473,755
Messages
2,569,536
Members
45,009
Latest member
GidgetGamb

Latest Threads

Top