isis-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stephen Cameron <steve.cameron...@gmail.com>
Subject Re: moving to test & production, how to optimise speed of response?
Date Tue, 29 Sep 2015 10:53:48 GMT
No measurement, just an observation. the reponse from the restful viewer
seems to me to be almost instantaneous, whereas the wicket viewer a few
seconds at least.

I am also comparing current Isis wicket performance against what I observe
in the security module video that Dan did, I'd say my speed is about half
what I see there.

Also, its slower than the system I am replacing, I cannot see that
difference being due to machine specs. That system is PHP, so template
based, its crap in other respects, but very fast. A OO server-side
component based system, I imagine, is always going to be bit slower due to
AJAX requests.


On Tue, Sep 29, 2015 at 8:42 PM, Jeroen van der Wal <jeroen@stromboli.it>
wrote:

> I've never seen the restful viewer outperform the wicket viewer. How did
> you measure?
>
> On 29 September 2015 at 12:25, Martin Grigorov <mgrigorov@apache.org>
> wrote:
>
> > On Tue, Sep 29, 2015 at 12:22 PM, Stephen Cameron <
> > steve.cameron.62@gmail.com> wrote:
> >
> > > I don't think its anything to do with database access, I am starting
> with
> > > an empty in-memory database. This will be an interesting challenge to
> try
> > > and track down the issue. I am struck by the big difference in
> > > responsiveness of the restful-objects viewer vs the wicket viewer, with
> > no
> > > criticism of wicket intended, all indications are that its performance
> is
> > > very good when benchmarked.
> > >
> >
> > Wicket is never the problem ;-)
> >
> > It should be easy for you to profile your app with jProfiler/YourKit.
> >
> >
> > >
> > > On Tue, Sep 29, 2015 at 2:06 AM, Dan Haywood <
> > dan@haywood-associates.co.uk
> > > >
> > > wrote:
> > >
> > > > There isn't really any list, no.
> > > >
> > > > Actually, on my todo list is to do some performance tuning; things
> have
> > > > certainly slowed down over the last few releases.  I have one or two
> > > > theories as to why, but haven't researched them as of yet.
> > > >
> > > > You might want to look configure logging to view all SQL queries and
> to
> > > > search for any occurrences of the N+1 problem; eg configure
> > > >
> > > > log4j.appender.sql=org.apache.log4j.RollingFileAppender
> > > > log4j.appender.sql.File=./logs/sql.log
> > > > log4j.appender.sql.Append=false
> > > > log4j.appender.sql.layout=org.apache.log4j.PatternLayout
> > > > log4j.appender.sql.layout.ConversionPattern=%d{yyyy-MM-dd
> HH:mm:ss.SSS}
> > > > %m%n
> > > >
> > > > and
> > > >
> > > > log4j.logger.DataNucleus.Datastore.Native=DEBUG, Console, sql
> > > > log4j.logger.DataNucleus.Datastore.Schema=DEBUG, Console, sql
> > > >
> > > > in logging.properties and then monitor sql.log.  I use baretail (on
> > > > windows) which supports highlighting of lines; makes it easy to see
> > which
> > > > tables are being hit repeatedly.
> > > >
> > > >
> > > > Another thing you could explore is to use selective caching, eg using
> > the
> > > > QueryResultsCache [1]
> > > >
> > > > HTH (a bit)
> > > > Dan
> > > >
> > > > [1]
> > > >
> > > >
> > >
> >
> http://isis.apache.org/guides/rg.html#_rg_services-api_manpage-QueryResultsCache
> > > >
> > > >
> > > >
> > > >
> > > > On 28 September 2015 at 13:51, Stephen Cameron <
> > > steve.cameron.62@gmail.com
> > > > >
> > > > wrote:
> > > >
> > > > > Hi,
> > > > >
> > > > > I am installing on a test server tomorrow, what can I set to
> optimise
> > > > speed
> > > > > of response? I was looking that the security module video online
> and
> > > > > thinking: gee that is fast! It would be good to get mine that nippy
> > :)
> > > > >
> > > > > I know that there are cache optimisations for Wicket and I see a
> > > warning
> > > > > that I am working in DEVELOPMENT mode when Jetty starts.
> > > > >
> > > > > Just wondering if there is a list anywhere of things to do in the
> > > > > switch-over?
> > > > >
> > > > > Cheers
> > > > > Steve C
> > > > >
> > > >
> > >
> >
>

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