On Thu, Sep 11, 2008 at 05:28:36PM +0200, Florian Lohoff wrote:
> Boinc, [EMAIL PROTECTED] and [EMAIL PROTECTED] show it. The work to be done 
> is always
> bigger than the cpu cycles available. And the more data the map will
> contain the longer the clients will take for a tile. So it is already
> foreseeable that the cpu cycles on the clients will not be enough.

I am not sure that is so. If oyu disregard the rerender of the world that is 
going on, we finish rerendering out current 6h cycle within 1-2 hours. Of 
course, if we were idling too much, we could simply reduce that time span and 
rerender every changed tile every hour, that would keep our clients more busy. 
But I ave doubts that this would provide much added value to what we have now.

I would rather see a stable system, where neither server nor client crash at 
the slightest of all problems, than one were all clients render useless tiles, 
just so they have something to render :-).

Right now, we can use all the CPU cycles we can get. And it's grat to have a 
big pool of clients that we can draw on when loads of things change, but I 
don't have problems with idling clients when there's nothing to do. I also 
don't have a problem with idling clients when the server processes 3k requests 
/ hour at the same time. :-) Admittedly, I have the server operators 
perspective, and not an end user one.

> I see the idling client to be a physcological problem. If you donate a
> CPU to [EMAIL PROTECTED] you want it to be used and feel not accepted when 
> the client
> idles.

That is a valid point. Client operators probably don't want to see the clients 
on the dedicated box idle.

_______________________________________________
Tilesathome mailing list
[email protected]
http://lists.openstreetmap.org/listinfo/tilesathome

Reply via email to