[keycloak-dev] Next step performance

[Stian Thorgersen]

Cache-control made a huge difference - halved the load time of the login screen!

Caching FreeMarker templates made much less of a difference (~10%). I'm a bit disappointed as I thought it would be a bigger win (it did load and parse 2-3 files templates for each page). 

----- Original Message -----
> From: "Stian Thorgersen" <stian at redhat.com>
> To: "Bill Burke" <bburke at redhat.com>
> Cc: keycloak-dev at lists.jboss.org
> Sent: Friday, 30 May, 2014 1:17:43 PM
> Subject: Re: [keycloak-dev] Next step performance
> 
> FIY I've started looking at caching FreeMarker templates and cache-control
> for static content
> 
> ----- Original Message -----
> > From: "Bill Burke" <bburke at redhat.com>
> > To: "Stian Thorgersen" <stian at redhat.com>
> > Cc: keycloak-dev at lists.jboss.org
> > Sent: Friday, 30 May, 2014 12:29:23 PM
> > Subject: Re: [keycloak-dev] Next step performance
> > 
> > 
> > 
> > On 5/30/2014 6:06 AM, Stian Thorgersen wrote:
> > > Sent invite for Monday in Zimbra (15pm GMT). If that's not suitable let
> > > me
> > > know.
> > >
> > > A few more areas to look at:
> > >
> > > * Profile to find crappy code
> > > * Tune FreeMarker and themes
> > > * Sharding / sticky sessions - this is probably for later, but I'd like
> > > to
> > > discuss it in the hangout
> > >
> > > A few comments on database caching:
> > >
> > > * We may need to do this in stages
> > >     - 2nd level cache for beta-2?
> > >     - Something more advanced in after 1.0.final
> > > * I'd don't like the idea of us re-inventing the wheel
> > >     - Caching has a lot of complex considerations (node discovery,
> > >     fail-over, node recovery, invalidation, lan/wan, jta, etc.)
> > 
> > I know and agree and have experience with this.  But, these traditional
> > caches just don't work well in many environments.
> > 
> > > * I really like the idea of having a cache model provider that delegates
> > > to
> > > another model for persistence / this is definitively the way to go IMO
> > > * User sessions needs to separate considerations, specifically I think we
> > > can come up with something clever to minimise the writes required
> > >
> > 
> > Yes, User sessions are a special case.
> > 
> > --
> > Bill Burke
> > JBoss, a division of Red Hat
> > http://bill.burkecentral.com
> > 
> _______________________________________________
> keycloak-dev mailing list
> keycloak-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/keycloak-dev
>