Re: Proposed module: tracker

[Jamie McCracken <jamiemcc blueyonder co uk>]

Gustavo J. A. M. Carneiro wrote:
>   Having tried tracker 0.5.3 for a couple of days what I have to say is:
>     1. it uses little memory;
>     2. it requires _a lot_ of CPU power.
>
>   Basically point 2 is a killer.  No one is going to want to run this
> except in servers.  Keeping the CPU busy almost 100% of the time is not
> nice: consumes more power, gets my computer fan running faster and more
> loudly.. I can't even begin to imagine what a nightmare it will be for
> laptop users..

point 2 is scheduled at nice +19 (same with Ionice +7) so it only uses 
more cpu if its idle. And cpu is only used if indexing (first run only 
takes a long time). After first run, cpu usage is generally negligible 
so its a one time hit.

Also try running a game or something cpu heavy while tracker is indexing 
and you will see tracker cpu drop to ~1%

I do believe this is the right way to schedule a service (via the OS) if 
not we can add periodic sleep periods but am not convinced we can do it 
better than the OS. I mean I can easily make it consume less than 50% of 
cpu by sleeping half the time but that will slow indexing proportionally.

Im open to people's thoughts on this... (IE is 100% cpu usage at 
nice+19/ionice+7 during indexing really a problem?)

(I note locate pretty much uses similar 100% cpu while indexing too)

>   Maybe this kind of indexing technology, be it tracker or beagle, is
> simply not something that we want to shove into users' desktops.  Either
> this gets much much better optimised in the future, or we have to wait
> for more powerful hardware.  In any case, I'm -1 for including tracker
> in GNOME 2.18; let's wait and see how this evolves at GNOME 2.20 time.

Im sure it will be better for 2.20. Release team can withdraw tracker 
from 2.18 proposed modules with my permission.


--
Mr Jamie McCracken
http://jamiemcc.livejournal.com/