I have 3-4 values.
First moskito gives me an min/average/max serving time for all my
servlets, actions and services. So I know whether the system really
got slower and where.
Second we measure cpu time spend in user mode and idle time and know
how much cpu time we consume per user.
In case user
Hi Leon,
you're right - performance is decreasing with log enabled, not increasing.
Would be nice if it was so ;-)
Unfortunately I don't know how to measure specific values, at the moment I'm
listening to my users ("it was faster yesterday without log enabled") and
colleagues ("I had a case wher
ops sorry, did i got you right?
my users are experiencing ___increasing___ performance if I enable access.log:
?
So enabling the log actually increases performance? Sounds like jrockit :-)
On 10/9/06, Frank Niedermann <[EMAIL PROTECTED]> wrote:
David,
CPU load is also very low, maximum is
do you have any resource problems?
How do you know that the performance is falling, and what does falling
exactly mean? 10%? 1%? 1 ms per request?
regards
Leon
On 10/9/06, Frank Niedermann <[EMAIL PROTECTED]> wrote:
David,
CPU load is also very low, maximum is 80%. There are two CPUs (real, n
David,
CPU load is also very low, maximum is 80%. There are two CPUs (real, not
virtualized) in the server and according to the performance view on Windows
there could be much more users on the system. But I'm not sure if that
performance view is true or not ...
Frank
David Smith-2 wrote:
>
>
But I'm not sure it would show as a disk bottleneck. If you have
frequent small writes to a disk and each write is delayed while
antivirus checks the datastream for virus signatures, the many tiny
delays could aggegate in to a much bigger file i/o slow down. The
system may experience a higher
You can have conditional access logging.
http://tomcat.apache.org/tomcat-5.0-doc/catalina/docs/api/org/apache/catalina/valves/AccessLogValve.html
condition='foo'
If set the Valve will look in the ServletRequest for an attribute called
foo. If it exists then the request is logged. (Of course you
David,
that is a good idea from far, far away :-)
Antivirus is enabled (I'm not suicidal, this is a Windows box ;) but
according to the Windows performance viewer there is no bottleneck on the
harddisk, it's always way under 10% load.
Frank
David Smith-2 wrote:
>
> I think a long time ago in
I think a long time ago in a galaxy far, far away I remember something
about antivirus impacting file I/O performance. Would your box happen
to have antivirus enabled? If so, any chance you could exclude your
logs from it and/or disable it for the purpose of a test?
--David
Frank Niedermann
Unfortunately I have to use Windows Server 2003 as the company behind the
application we're using is not supporting UNIX/Linux.
Windows also has performance utilities but they tell me that the server
isn't heavily loaded at all.
A good think would be to have a smaller access log just for statist
Something seems odd with your system. I have pounded some tomcat
installations with old unix hardware with and without access logging and
could hardly tell the difference.
In linux - i was able to tell more of a difference, but not enough to
turn off logging.
I am at a loss of where the bott
Sorry for the strange quoting, that was me using nabble.com ...
My reply should also contain (but does not - don't know what went wrong):
I've installed LambdaProbe and it tells me that there are not much Threads
(about 50) and most of them are in state of waiting or timed_waiting. So
that seem
Tim,
Tim Funk wrote:
>
> Unless you are max'd on working threads - access logging should not be a
> performance hit. Access logging takes pace after the response is sent to
> the client.
>
BUT if the access logs are big, AND you a re low on disk, AND/OR your
disk is SLW then that could
Unless you are max'd on working threads - access logging should not be a
performance hit. Access logging takes pace after the response is sent to
the client.
BUT if the access logs are big, AND you a re low on disk, AND/OR your
disk is SLW then that could be a problem. The overhead of log
14 matches
Mail list logo
