You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@turbine.apache.org by Kurt Schrader <ks...@engin.umich.edu> on 2001/07/16 22:18:53 UTC

Updated Turbine 3.0 TDK Benchmarks

Using the current version of the Turbine code, Turbine has shown even 
more improvement in 3.0 than it did the last time I posted benchmarks.  
Currently, with concurrency at 20 connections and 500 connections total 
in Apache Bench I get the following measurements on a fresh version of 
the TDK after in initial click thru of the pages to make sure that the 
initialization time of the servlet wouldn't play a role:

Turbine 2.1 (release version)
--------------------------------------------------------------
Connnection Times (ms)
                         min   avg   max
Connect:             0     3   118
Processing:    231   660  1359
Total:               231   663  1477

Turbine 3.0 (Current Alpha code base)
--------------------------------------------------------------
Connnection Times (ms)
                        min   avg   max
Connect:           0     9    70
Processing:      3    84   522
Total:                 3    93   592

Using the average times, we already see an improvement of a factor 7+ 
over 2.1 with no performance tuning as of yet.  3.0 is looking to be an 
awesome improvement.

-Kurt

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


Re: Updated Turbine 3.0 TDK Benchmarks

Posted by Kurt Schrader <ks...@engin.umich.edu>.
On Tuesday, July 17, 2001, at 01:24 AM, fanyun wrote:

 > How many memory do you have on the Mac

256 megs

On Tuesday, July 17, 2001, at 02:23 PM, Jason van Zyl wrote:

> <cut>
>
> Kurts benchmarks are a non-caching 2.1 against a non-caching 3.0.
>

Caching 2.1 results
------------------------------------------------------------
Connnection Times (ms)
                        min   avg   max
Connect:           0     2    27
Processing:      4   105   824
Total:                 4   107   851

Much faster than non-caching 2.1, about on par with my 3.0 results.

-Kurt

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


Re: Updated Turbine 3.0 TDK Benchmarks

Posted by Jason van Zyl <jv...@apache.org>.
On 7/17/01 2:40 AM, "Ilkka Priha" <il...@simsoft.fi> wrote:

> I know there is a lot of redundancy in the page building process of 2.1, but
> that big difference compared to the new one is a surprise. Typically such
> differences are only caused by larger number of i/o operations, which have
> been eliminated (at least have tried to be) in 2.1 also. Actually, I was
> expecting 2.1 to be still slightly faster as it caches both the results of
> module loadings and template searches while the initial version of the new
> Resolver doesn't yet cache its results.
> 
> Just to make sure, have you set the module.cache property to true in TR
> (default is false)? The caching of module loaders and template searches in
> 2.1 is controlled by that property; 2.2 ModuleLoader caches always and 2.2
> Resolver doesn't use a cache yet.

There is no caching at all in 3.0. The ModuleLoader is not caching at all.
You can verify this yourself by running the sample TDK app and setting the
priority for the default category to DEBUG. There are debug statements
when a module is searched for and a debug statement when a module is
placed in the cache. The sample app in the 3.0 TDK shows the continued
searching for the same module and no caching.

Kurts benchmarks are a non-caching 2.1 against a non-caching 3.0.
 
> -- Ilkka
> 
> 
> ----- Original Message -----
> From: "Kurt Schrader" <ks...@engin.umich.edu>
> To: <tu...@jakarta.apache.org>
> Sent: Tuesday, July 17, 2001 07:48
> Subject: Re: Updated Turbine 3.0 TDK Benchmarks
> 
> 
>> 
>> On Monday, July 16, 2001, at 09:34 PM, fanyun wrote:
>> 
>>> Can you tell us which kind of machine do you do the test.
>>> 
>> 
>> Mac OS X, 500MHz G4, Apple JDK 1.3.1 Beta
>> 
>> I have a Linux box and a Wintel Box that I could grab some
>> comparative benchmarks on tomorrow if people are
>> interested in that sort of thing.
>> 
>> -Kurt
>> 
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: turbine-dev-unsubscribe@jakarta.apache.org
>> For additional commands, e-mail: turbine-dev-help@jakarta.apache.org
>> 
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: turbine-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: turbine-dev-help@jakarta.apache.org

-- 

jvz.

Jason van Zyl

http://tambora.zenplex.org
http://jakarta.apache.org/turbine
http://jakarta.apache.org/velocity
http://jakarta.apache.org/alexandria
http://jakarta.apache.org/commons



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


Re: Updated Turbine 3.0 TDK Benchmarks

Posted by Ilkka Priha <il...@simsoft.fi>.
I know there is a lot of redundancy in the page building process of 2.1, but
that big difference compared to the new one is a surprise. Typically such
differences are only caused by larger number of i/o operations, which have
been eliminated (at least have tried to be) in 2.1 also. Actually, I was
expecting 2.1 to be still slightly faster as it caches both the results of
module loadings and template searches while the initial version of the new
Resolver doesn't yet cache its results.

Just to make sure, have you set the module.cache property to true in TR
(default is false)? The caching of module loaders and template searches in
2.1 is controlled by that property; 2.2 ModuleLoader caches always and 2.2
Resolver doesn't use a cache yet.

-- Ilkka


----- Original Message -----
From: "Kurt Schrader" <ks...@engin.umich.edu>
To: <tu...@jakarta.apache.org>
Sent: Tuesday, July 17, 2001 07:48
Subject: Re: Updated Turbine 3.0 TDK Benchmarks


>
> On Monday, July 16, 2001, at 09:34 PM, fanyun wrote:
>
> > Can you tell us which kind of machine do you do the test.
> >
>
> Mac OS X, 500MHz G4, Apple JDK 1.3.1 Beta
>
> I have a Linux box and a Wintel Box that I could grab some
> comparative benchmarks on tomorrow if people are
> interested in that sort of thing.
>
> -Kurt
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: turbine-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: turbine-dev-help@jakarta.apache.org
>


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


RE: Updated Turbine 3.0 TDK Benchmarks

Posted by fanyun <yu...@vtradex.com>.
How many memory do you have on the Mac


Regards

fanyun




-----Original Message-----
From: Kurt Schrader [mailto:kschrade@engin.umich.edu]
Sent: Tuesday, July 17, 2001 12:49 PM
To: turbine-dev@jakarta.apache.org
Subject: Re: Updated Turbine 3.0 TDK Benchmarks



On Monday, July 16, 2001, at 09:34 PM, fanyun wrote:

> Can you tell us which kind of machine do you do the test.
>

Mac OS X, 500MHz G4, Apple JDK 1.3.1 Beta

I have a Linux box and a Wintel Box that I could grab some
comparative benchmarks on tomorrow if people are
interested in that sort of thing.

-Kurt

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


Re: Updated Turbine 3.0 TDK Benchmarks

Posted by Jon Stevens <jo...@latchkey.com>.
on 7/16/01 9:48 PM, "Kurt Schrader" <ks...@engin.umich.edu> wrote:

> Mac OS X, 500MHz G4, Apple JDK 1.3.1 Beta
> 
> I have a Linux box and a Wintel Box that I could grab some
> comparative benchmarks on tomorrow if people are
> interested in that sort of thing.
> 
> -Kurt

Yea, and that JVM is pitifully slow compared with other JVM's on other
platforms. :-)

-jon


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


Re: Updated Turbine 3.0 TDK Benchmarks

Posted by Kurt Schrader <ks...@engin.umich.edu>.
On Monday, July 16, 2001, at 09:34 PM, fanyun wrote:

> Can you tell us which kind of machine do you do the test.
>

Mac OS X, 500MHz G4, Apple JDK 1.3.1 Beta

I have a Linux box and a Wintel Box that I could grab some
comparative benchmarks on tomorrow if people are
interested in that sort of thing.

-Kurt

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


RE: Updated Turbine 3.0 TDK Benchmarks

Posted by fanyun <yu...@vtradex.com>.
Can you tell us which kind of machine do you do the test.


Regards

fanyun



-----Original Message-----
From: Kurt Schrader [mailto:kschrade@engin.umich.edu]
Sent: Tuesday, July 17, 2001 4:19 AM
To: turbine-dev@jakarta.apache.org
Subject: Updated Turbine 3.0 TDK Benchmarks


Using the current version of the Turbine code, Turbine has shown even 
more improvement in 3.0 than it did the last time I posted benchmarks.  
Currently, with concurrency at 20 connections and 500 connections total 
in Apache Bench I get the following measurements on a fresh version of 
the TDK after in initial click thru of the pages to make sure that the 
initialization time of the servlet wouldn't play a role:

Turbine 2.1 (release version)
--------------------------------------------------------------
Connnection Times (ms)
                         min   avg   max
Connect:             0     3   118
Processing:    231   660  1359
Total:               231   663  1477

Turbine 3.0 (Current Alpha code base)
--------------------------------------------------------------
Connnection Times (ms)
                        min   avg   max
Connect:           0     9    70
Processing:      3    84   522
Total:                 3    93   592

Using the average times, we already see an improvement of a factor 7+ 
over 2.1 with no performance tuning as of yet.  3.0 is looking to be an 
awesome improvement.

-Kurt

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