[Tilesathome] Long Request Take Time?

[Matthias Julius]

Dirk-Lüder Kreie <osm-list at deelkar.net> writes:

> I'm going to elaborate a bit on what I wrote in this thread already.
>
> Ian Dees schrieb:
>> But this means my CPUs are sitting idle with no work to do, and I'm paying
>> for their power and heat to do nothing. I'd rather see them doing preemptive
>> renders of highly-edited areas or rendering those prio4 requests than
>> sitting idle.
>
> Basically, tiles at home is not like other distributed computing
> projects. Other projects solve workunits, and they remain solved. The
> base data does not change, but at most gets amended, and thus creates
> new workunits. After a workunit is done and returned to the server, it
> needs very little work on the server side. All this allows other
> distributed projects to almost always run at full speed.
>
> T at H, on the other hand, has a limited workunit space of about 17
> million workunits that have to be almost randomly, but constantly,
> updated, so you can't predict just how much crunching power you'll
> need in the next X hours, even without those ultra-easy prio4 tilesets
> we seem to render so much of, lately.

And without those prio4 requests there simply wouldn't be any work to
do for the clients at times.

There are probably many areas in the US that have not seen any edits
since the TIGER import.  Once the prio4 requests get into those areas
the clients will get real work to do.

Matthias