Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

General J2EE and database use performance issue questions...

1 view
Skip to first unread message

Valentino Kyriakides

unread,
Nov 8, 2001, 12:59:19 PM11/8/01
to
Hi,

we've build some internal J2EE test apps just to do some conceptual
requirements testings for some possible upcoming project. We used some
container managed EJBs to fetch out 1000 address records (which are crowded
amoung different field, value, metadata etc. tables) from a MySQL database.
The so far results were:

- for Sun's reference J2EE implementation, fetching 1000 addresses it took
~330 seconds :-(
- for jBoss, fetching 1000 addresses it took ~120 seconds

- for jBoss using the container caching mechanism for fetching the 1000
addresses it always took ~108 seconds :-(
- for direct JDBC usage (no EJBs at all) for fetching out the 1000 addresses
it took ~6 seconds :-)
- note that we didn't tested the above mentioned with BAS, WebSphere or BEA
so far.

Well, the problem is, that we might want to use EJBs (an J2EE appserver) but
the above shown so far results are not very motivating, since we would need
to be able to get out at least 100000 addresses in some resonable time from
a RDBMS. As the above test scenario shows, only the direct JDBC calls are so
far somehow acceptable.

The general questions I have are related to possible J2EE speed
optimizations.

Would the usage of EJB-2.0 specific local interfaces and local home
interfaces dramatically speed up the EJB processing?

Would the usage of bean managed persistence entity beans offer also more
speed beside more flexibility in accessing a RDBMS?

Would the usage of direct object serialisation writing/reading of address
objects into ...say... JDataStore (instead of using JDBC and RDBMS like
tables) offer more speed and better timing results? Does or is this sort of
object persisteny via JDataStore usable via EJBs?


Generally I would be interested in some optimization recommendations and
comments, which could improve the above sketched J2EE szenario.


Thanks Valentino

Valentino Kyriakides

unread,
Nov 10, 2001, 6:16:49 AM11/10/01
to
Another test with BMP instead of CMP resulted now to 13 sec instead of 120
sec for retrieving 1000 addresses with Sun's reference App.Server. So I
believe that other app servers like jBoss, BAS, etc. will even decrease
these 13 secs to some better timings.

Valentino

"Valentino Kyriakides" <vk...@ision.net> schrieb im Newsbeitrag
news:3beac559_1@dnews...

Chuck Davis

unread,
Nov 11, 2001, 5:00:13 PM11/11/01
to
Valentino:

I've read that it will make a substantial difference. Perhaps close to
what you are getting with BMP. I am not experienced but what I read
indicates EJB2.0 with local interfaces is really the entry-level product as
far as usability is concerned. Have to avoid remote interfaces because
they go through the TCP/IP stack even if running on the same machine becaue
they are running in differenct JVMs.

Hopefully some more knowledgable, experienced person will pop by soon with
real answers.

Chuck Davis

0 new messages