Re: Desktop search: should background processes be (io)niced?
On Monday 08 March 2010, Valerio Passini wrote:
> Alle lunedì 08 marzo 2010, Michael Schuerig ha scritto:
> > I'm trying to take advantage of the newly usable desktop search
> > feature and have selected several GBs of documents for indexing.
> > The drawback is that the various processes associated with this
> > task -- nepomukservices, virtuoso-t -- chew heavily on the CPU and
> > slurp away I/O bandwidth.
> >
> > I noticed one nepomukservices process that is running at nice level
> > 19, all other related processes apparently have nice level 0. I
> > don't know if there is any ionice-ing done at all. Shouldn't all
> > these processes be running in such a way as to yield CPU and I/O
> > bandwidth to other processes?
> I was as willing as you to try it, but there is problem with strigi
> starting indexing the same stuff over and over again. See bug:
> https://bugs.kde.org/show_bug.cgi?id=226895
> In theory the desktop search engine should eat much CPU only once, at
> the beginning, when the index must be built from zero. When this
> process is completed, only newer files need to be indexed and the
> CPU usage is comparably lower. If you don't meet that bug, you
> probably should be happy with nepomuk.
I'm not convinced this is a real bug. From what I can tell, strigi does
not index in depth-first order. In my case, I have >> 100.000 files that
I'd like to have indexed (if that's really sensible, I'll decide later)
and I'm prepared for it to take some time. But still I don't think
indexing should hog my computer the way it does.
Michael
--
Michael Schuerig
mailto:michael@schuerig.de
http://www.schuerig.de/michael/
Reply to: