tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sérgio Costa" <oscil...@gmail.com>
Subject Re: Tomcat's scalability
Date Tue, 20 Jun 2006 16:39:08 GMT
Be aware that Firewalls are known to drop idle TCP conections after a
certain amount of time.

We experienced a similar problem with a DB2 JDBC conection pool through a
Cisco PIX.

What was happening was the conections were closed by the firewall. When they
were re-used, there was an added overhead until Java noticed the socket was
closed, and the reconected it.

We solved it through firewall configuration.

Sérgio

On 6/20/06, Hyatt, Gordon <Gordon.Hyatt@joslin.harvard.edu> wrote:
>
> We had an application (with connection pooling and validation query)
> running on TC5.0.28 (now on TC5.5.17) connecting to MS SQL Server using
> the Microsoft JDBC drivers.  When tomcat and the database were on the same
> side of the firewall performance was great (averaged < 20ms obtaining
> connection and 2-3s running SELECT queries).  (Yes, I actually put tracing
> into the code that timed how long it took to acquire the connection, run the
> query, and transmit the data back to the client.)
>
> Once we moved the tomcat server outside the firewall (an the SQL Server
> stayed inside), the performance degraded horribly (7-10s to obtain
> connection, 15-20s to run the simple queries (select by primary key), > 30s
> to run the more 'complex' queries, for example SELECT id, name, description
> FROM single_table WHERE name LIKE '%a%';).
>
> As a test, I changed the database to PostgreSQL v8.0 (at that time it was
> still in beta), changed the driver, and ported the database (dropped all
> indices except primary key and foreign keys), the query results were
> dramatically different: < 500ms on any query I tried.  I had 2 instanced of
> my app on the tomcat server, one using PostgreSQL, the other using MS SQL
> Server.  A member of the Network Team sat with me and opened both ports
> (1433 and 5432) to the database machine.
>
> I even reconfigured PostgreSQL to use the same port as SQL Server,
> restarted tomcat, ran a set of queries using one DB then the other (shutting
> the first down, obviously).  The timing was still as dramatically different
> as running on 2 different ports.
>
> I know there may have been configuration issues with the firewall that
> impeded the performance, but, unfortunately, that (the firewall
> configuration) was beyond my control.
>
>
> Basically, be warned: if you are running your connection to MS SQL Server
> (and using the Microsoft JDBC drivers) through a firewall, this MAY be one
> contributing factor of your poor performance.
>
> Just my $0.02.
>
> Gord
>
>
> -----Original Message-----
> From: Alex Turner [mailto:armtuk@gmail.com]
> Sent: Tuesday, June 20, 2006 12:42 AM
> To: Tomcat Users List
> Subject: Re: Tomcat's scalability
>
>
> On 6/19/06, Leon Rosenberg <rosenberg.leon@googlemail.com> wrote:
> >
> > On 6/19/06, Biernatowski Bartosz J <Bartosz.Biernatowski@rxsol.com>
> wrote:
> > > I am about 90% sure the bottleneck is Tomcat or what's running on top
> of
> > > Tomcat. Application uses JDBC queries to MS SQL server
> > > Chips are Intel Xeon. My monitoring data:
> > > Memory utilization under 30%, CPU under 10%. Using hardcore
> performance
> > > tools and systematic approach.
> > > The bottom line is that Tomcat/my application combo don't seem to
> handle
> > > more than a certain number of users. All I want to do is to up the #
> of
> > > users by 3.
> >
> > Sounds like your db connection pool is the problem. Maybe you should
> > check whether you have enough connections in the connection pool.
>
>
> Connection pooling, ahh yes.  This is also a likely problem if you aren't
> doing any.
>
>
> >
> > > So far it sounds that the approach of adding separate instance of
> Tomcat
> > and
> > > using round robin is better than adding a separate JVM.
> > I think both options are equal. How do you plan to run a separate
> > tomcat in the same JVM?
>
>
> Both options are equaly stupid.  Putting multiple intances of tomcat on a
> single box is pretty worthless unless you have serious application
> problems.  Tomcat is multi-threaded, and will by nature utilize a
> multi-CPU
> setup.
>
>
> If you ask me (and hey, we have thousands of concurrent users and a
> > lot more requests) you need a monitoring tool for your application
> > inside your application not just vmstat or top. You need to know which
> > servlet/action/whatever your presentation layer is takes the time and
> > trace it down in the persistence. Everything else is just kindergarten
> > :-)
> >
> >
> > >
> > >
> > > BJ Biernatowski
> > > Application Developer, e-Business
> >
> > Leon
> >
> > >
> > > -----Original Message-----
> > > From: Leon Rosenberg [mailto:rosenberg.leon@googlemail.com]
> > > Sent: Monday, June 19, 2006 10:49 AM
> > > To: Tomcat Users List
> > > Subject: Re: Tomcat's scalability
> > >
> > > are you sure that tomcat is your bottleneck?
> > > Your 4 CPU machine (which cpu's btw?) should be able to handle more
> > > than 1000 users (unless you are speaking about suns cpu) without
> > > problems. Maybe you should provide more info about your application.
> > > Do you have any monitoring data?
> > >
> > > Leon
> > >
> > > On 6/19/06, Biernatowski Bartosz J <Bartosz.Biernatowski@rxsol.com>
> > wrote:
> > > > Hello,
> > > > I was hoping somebody on the list might point me in the right
> > direction...
> > > >
> > > > I am trying to scale up Tomcat based web application currently
> > supporting
> > > > ~100 users to 350 users.
> > > >
> > > > It seems that I have enough hardware: 2 load balanced servers x 4
> CPUs
> > > each
> > > > with 4 GB of RAM which is underutilized for most of the time even
> > though
> > > > application performance slows dramatically at peak times.
> > > >
> > > > I was advised to install multiple JVMs in order to improve Tomcat's
> > > > performance. Another option I considered was to install 2 instances
> > > > of Tomcat on each server to see whether it would handle increased
> > load.
> > > >
> > > > Would anybody know what kind of performance improvement would
> multiple
> > > > JVM/Tomcat installations provide? Are there any benchmarks
> available?
> > > >
> > > > Thank you for any help!
> > > > BJ
> > > >
> > > > BJ Biernatowski
> > > > Application Developer
> > > >
> > > > This e-mail, including attachments, may include confidential and/or
> > > proprietary information, and may be used only by the person or entity
> to
> > > which it is addressed. If the reader of this e-mail is not the
> intended
> > > recipient or his or her authorized agent, the reader is hereby
> notified
> > that
> > > any dissemination, distribution or copying of this e-mail, including
> its
> > > contents and attachments, is prohibited. If you have received this
> > e-mail in
> > > error, please notify the sender by a "reply to sender only" message
> and
> > > delete this e-mail immediately and destroy all electronic and hard
> > copies of
> > > the communication, including attachments.
> > > >
> > > >
> > > >
> ---------------------------------------------------------------------
> > > > To start a new topic, e-mail: users@tomcat.apache.org
> > > > To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> > > > For additional commands, e-mail: users-help@tomcat.apache.org
> > > >
> > > >
> > >
> > > ---------------------------------------------------------------------
> > > To start a new topic, e-mail: users@tomcat.apache.org
> > > To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> > > For additional commands, e-mail: users-help@tomcat.apache.org
> > >
> > >
> > > This e-mail, including attachments, may include confidential and/or
> > proprietary information, and may be used only by the person or entity to
> > which it is addressed. If the reader of this e-mail is not the intended
> > recipient or his or her authorized agent, the reader is hereby notified
> that
> > any dissemination, distribution or copying of this e-mail, including its
> > contents and attachments, is prohibited. If you have received this
> e-mail in
> > error, please notify the sender by a "reply to sender only" message and
> > delete this e-mail immediately and destroy all electronic and hard
> copies of
> > the communication, including attachments.
> > >
> > >
> > > ---------------------------------------------------------------------
> > > To start a new topic, e-mail: users@tomcat.apache.org
> > > To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> > > For additional commands, e-mail: users-help@tomcat.apache.org
> > >
> > >
> >
> > ---------------------------------------------------------------------
> > To start a new topic, e-mail: users@tomcat.apache.org
> > To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> > For additional commands, e-mail: users-help@tomcat.apache.org
> >
> >
>
> ---------------------------------------------------------------------
> To start a new topic, e-mail: users@tomcat.apache.org
> To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> For additional commands, e-mail: users-help@tomcat.apache.org
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message