Hi Toni,
I had a quick look at the config files, I would recommend trying
WaitForCompletion policy="*first*" instead of WaitForCompletion
policy="*all*".
You seem to have added a lot of indexes on the main recovery log table,
this will significantly slowdown inserts in that table. As this table is
mostly written than read, reducing the number of indexes should give you
a better balance in performance.
The other settings seem fine. It looks like you are loading 3 other
virtual databases in your controller.xml file, make sure the settings
are similar.
Also make sure that you don't have too much logging in log4j.properties
as this can significantly slowdown the system if you turn some of them
to DEBUG.
Hope this helps,
Emmanuel
Sorry I think I missed the configuration files, can you have a look
just to see if you find something wrong in them, or simething that I
can improve !
2009/3/20 Toni Menendez <tonime...@gmail.com <mailto:tonime...@gmail.com>>
First of all,
Thanks a lot for the help, now I write down responses ?
Emmanuel Cecchet escribió:
Toni Menendez Lopez wrote:
How many controllers are you using?
>>I am using 2 Back-Ends, which on controller in each-backend.
I did not get that. You have 2 databases on 1 controller?
I have 4 databases in one controller, and 1 controller on each BE.
Which version of MySQL are you using? >>server version:
4.1.20-log
What components are located where? >>What do you mean
with this ?
Do you have the controller(s) and the databases on different
machines?
Yes, I have 2 BackEnds on each backend same databases doing
replication.
Is your JBoss server on another machine?
No, same servers !
What is your network architecture (how are they interconnected)?
I have a dedicated switch for this machines, so are connected in
same network witn 100Mb ethernet configuration.
Did you co-locate JBoss and Sequoia?
What do you mean with this ? You mean if they are in the same
server ? In that case : yes.
Which operating system are you using? >>Redhat Enterprose
Linus 4 update 4 Which JVM are you using?
>>java -version
>>java version "1.5.0_10"
>>Java(TM) 2 Runtime Environment, Standard Edition (build
1.5.0_10-b03)
>>Java HotSpot(TM) Server VM (build 1.5.0_10-b03, mixed mode)
What is your network configuration? >>Network
configuration, I have a private network to connect the 2
BE between each one, and after we are using jgroups.
If you have a single controller, you should not use group
communication and see big benefits.
I have 2 controllers one for each server !
If you have 2 controllers (one on each backend), you should
try different JGroups settings. Oftentimes, the TCP stack
gives better results. You can also give it a try with Appia to
see if performance improves.
I send you my configuration files, to see what do you think about.
Sorry, how can I use Appia ?
What does your workload look like (reads vs writes,
transactions, ...)? >>In our case mainly writes !
Well, then you might want to reconsider the choice of Sequoia.
Sequoia will not provide you speedup for writes (just for
reads) and will certainly significantly slowdown your write
throughput. If you don't have strong consistency constraints,
you might want to consider using asynchronous replication
instead like MySQL replication.
We used sequoia mainly becouse the high availability solution for
mysql.
So, you are telling me that in case of too much writes/sec sequoia
is not a good solution ? Why ?
In our application we need both lots of reads and lots of writes.
Is there any way to write in mysql and this changes can be read
later by sequoia ?
How many writes/sec do you think we can achive in sequoia with a
very simple INSERT ?
What do you mean using asyncronous replication ?
If you are using a single controller, you might want to set
you WaitForCompletion policy to 'first' in the load balancer
settings of your virtual database configuration file.
Hope this helps,
Emmanuel
Are you using direct JDBC or Hibernate? >>Hibernate.
These are preliminary questions that should help
highlighting
points worth investigating. Tuning requires some
analysis of your
system to understand what the potential bottlenecks can
be. This
is specific to your system and you might want to look
for someone
providing consulting services in your area for an
appropriate
answer to your needs
>> First of all I would like to try to do the tinning by
ourselves before contacting any external consulting. I
would like to know if there is any parameter to tune the
sequoia.
Thanks for your interest in Sequoia,
Emmanuel
-- Emmanuel Cecchet
FTO @ Frog Thinker Open Source Development & Consulting
--
Web: http://www.frogthinker.org
<http://www.frogthinker.org/> <http://www.frogthinker.org/>
email: m...@frogthinker.org
<mailto:m...@frogthinker.org> <mailto:m...@frogthinker.org
<mailto:m...@frogthinker.org>>
Skype: emmanuel_cecchet
_______________________________________________
Sequoia mailing list
Sequoia@lists.forge.continuent.org
<mailto:Sequoia@lists.forge.continuent.org>
<mailto:Sequoia@lists.forge.continuent.org
<mailto:Sequoia@lists.forge.continuent.org>>
https://forge.continuent.org/mailman/listinfo/sequoia
------------------------------------------------------------------------
_______________________________________________
Sequoia mailing list
Sequoia@lists.forge.continuent.org
<mailto:Sequoia@lists.forge.continuent.org>
https://forge.continuent.org/mailman/listinfo/sequoia
Thanks,
Toni.
------------------------------------------------------------------------
_______________________________________________
Sequoia mailing list
Sequoia@lists.forge.continuent.org
https://forge.continuent.org/mailman/listinfo/sequoia
--
Emmanuel Cecchet
FTO @ Frog Thinker
Open Source Development & Consulting
--
Web: http://www.frogthinker.org
email: m...@frogthinker.org
Skype: emmanuel_cecchet
_______________________________________________
Sequoia mailing list
Sequoia@lists.forge.continuent.org
https://forge.continuent.org/mailman/listinfo/sequoia
