[jira] Commented: (WICKET-3091) Remove references to last page from SecondLevelCacheSessionStore from idle sessions to reduce memory usage

["Igor Vaynberg (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/WICKET-3091?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12920676#action_12920676 ] 

Igor Vaynberg commented on WICKET-3091:
---------------------------------------

there isnt much point reducing 170 megs of memory down to 60. 170 megs for 1500 concurrent sessions is pretty damn good. i'd understand the concern if we were getting close to some sort of a limit, but 170 megs in modern systems is nothing. sometimes my twitter client takes up more :)

that said, a possible solution would be to enable passivation of idle sessions in the container. this way the container itself will serialize the session to disk when it is idle. problem solved.

> Remove references to last page from SecondLevelCacheSessionStore from idle sessions to reduce memory usage
> ----------------------------------------------------------------------------------------------------------
>
>                 Key: WICKET-3091
>                 URL: https://issues.apache.org/jira/browse/WICKET-3091
>             Project: Wicket
>          Issue Type: Improvement
>          Components: wicket
>    Affects Versions: 1.4.12
>            Reporter: Stefan Fussenegger
>
> I've already described this idea in WICKET-2889. The proposed implementation (patch attached to WICKET-2889) was rejected though:
> > From: Johan Compagner <jc...@gmail.com> 
> > Date: Mon, 9 Aug 2010 17:00:12 +0200 
> > Message-ID: <AA...@mail.gmail.com> 
> > Subject: Re: Could somebody please comment on WICKET-2889? 
> > To: users@wicket.apache.org 
> > Content-Type: text/plain; charset=ISO-8859-1 
> > 
> > I am against that patch. 
> > You keep pagemaps in memory in the applicaiton context! 
> > Those are session stuff stored in the HttpSession. You shouldnt keep 
> > reference to those stuff. 
> > This can break all kind of things (for example clustering) 
> > 
> > If you want something like that, then it is fine if we need to change 
> > something so that you can overwrite some stuff so that you can do what 
> > you want without patching wicket. 
> > But i dont want this to be default in wicket. 
> As I understood, the objects are to the implementation not the idea itself. I still consider this a very valuable improvement as it has the potential to drastically reduce Wicket's memory consumption in productions environments, especially for applications that make (heavy) use of statefull pages. An average page in our application used 115K of memory. With 1500 live sessions within 30 minutes, last pages accumulated for 170M memory usage. By expiring the last page after 10 minutes, this could be cut by 2/3 or 113M. (btw: if I remember correctly, Wicket's heap usage wasn't more than 220M in total, hence reducing overall memory usage by about 50%)

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.