Re[4]: Tests of mail list.

[Eugene Drobitko <eu...@luciddream.com>]

You wrote 30 мая 2003 г., 19:19:20:

>> I see that speed for list about 1000-2000 users is bigger.
I>> f first to process the first part then the second and etc.  then the mean
>> speed will be the same as for each part. The same for memory.

NJB> With respect to memory, you are mistaken.  The location where the memory is
NJB> lost is somewhat surprising.  sqlResources strings appear to be taking up
NJB> all of the memory.  There is a separate copy of them for each user mailbox.
NJB> The amount of space is directly related to the total number of mailboxes
NJB> instantiated, not the number per message.
OK.

NJB> With respect to getUserByName performance, I believe that you are also
NJB> mistaken, but there is a test you can try.  Put 25,000 users in the user
NJB> repository, but put only 2500 in the list database.  If the recipient list
NJB> matters, the performance will be 10 messages per second.  If the size of the
NJB> local repository matters, the performance will be much slower.
You're right.
I re-tested for 25,000 users in user repository and 2500 in mail list.
The delivery speed is less in 3 times.

>> That version of MySQL do You use?

NJB> 3.23.42 on the test server.

NJB>         --- Noel


NJB> ---------------------------------------------------------------------
NJB> To unsubscribe, e-mail: james-dev-unsubscribe@jakarta.apache.org
NJB> For additional commands, e-mail: james-dev-help@jakarta.apache.org




Best regards, Eugene                          mailto:eugene@luciddream.com


---------------------------------------------------------------------
To unsubscribe, e-mail: james-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: james-dev-help@jakarta.apache.org

RE: Tests of mail list.

["Noel J. Bergman" <no...@devtech.com>]

So far I have been able to improve the memory situation by inserting

    if ( dbMatcherElement != null ) {
        dbProduct = matchDbConnection(conn, dbMatcherElement);
+       m_perl5Util = null;     // release the PERL matcher!
    }

in SQLResources.java, which frees up 1000s of no longer needed PERL
matchers.  That allowed me to get up to ~8800 mailboxes before running out
of memory.  But the big memory is still taken up by the SQL strings.

> You're right.
> I re-tested for 25,000 users in user repository and 2500 in mail list.
> The delivery speed is less in 3 times.

I've put some code into getUserByName to see where it is spending the time.

	--- Noel


---------------------------------------------------------------------
To unsubscribe, e-mail: james-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: james-dev-help@jakarta.apache.org