From kde-panel-devel Tue May 28 09:43:00 2013 From: Vishesh Handa Date: Tue, 28 May 2013 09:43:00 +0000 To: kde-panel-devel Subject: Re: Activities - Storing ResourceScoreCache in Nepomuk Message-Id: X-MARC-Message: https://marc.info/?l=kde-panel-devel&m=136973347430864 MIME-Version: 1 Content-Type: multipart/mixed; boundary="--===============8062227628940630255==" --===============8062227628940630255== Content-Type: multipart/alternative; boundary=001a11c3702c5fa5af04ddc3ec2f --001a11c3702c5fa5af04ddc3ec2f Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable On Tue, May 28, 2013 at 2:44 PM, Ivan =C4=8Cuki=C4=87 = wrote: > > > I'm obviously still in favour of scrapping the code or at least disabli= ng > > it for 4.11 > > I'm leaning towards that decision as well - disabling until we start real= ly > using it somewhere. > Do you want me to provide a patch? > One thing I wanted to ask you - did you do some benchmarks about how much > virtuoso slows down when the database gets twice as big? (from my pov, if > the > performance is drastically changed, it means either that the tables are n= ot > properly indexed or that virtuoso is crap :) ) > Virtuoso is meant to scale so I doubt the performance will be that bad. It depends on the kind of query. The main problem is the lack of indexes for literal terms. Since all the "objects" in a statement are stored in the same column, we cannot have specialized indexes for dates, integers, etc. So when doing a search through for a certain date range and other similar stuff, it is doing a manual scan. Also, virtuoso never removes its indexes. It only updates them, cause in the "semantic word" deletions are very rare. > Cheerio, > Ivan > > > -- > Money can't buy happiness, but neither can poverty. > -- Leo Rosten > > _______________________________________________ > Plasma-devel mailing list > Plasma-devel@kde.org > https://mail.kde.org/mailman/listinfo/plasma-devel > > --=20 Vishesh Handa --001a11c3702c5fa5af04ddc3ec2f Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable



On Tue, May 28, 2013 at 2:44 PM, Ivan =C4=8Cuki=C4=87 <ivan.cukic= @kde.org> wrote:

> I'm obviously still in favour of scrapping the code or at least di= sabling
> it for 4.11

I'm leaning towards that decision as well - disabling until we st= art really
using it somewhere.

Do you want me to p= rovide a patch?


One thing I wanted to ask you - did you do some benchmarks about how much virtuoso slows down when the database gets twice as big? (from my pov, if t= he
performance is drastically changed, it means either that the tables are not=
properly indexed or that virtuoso is crap :) )

Virtuoso is meant to scale so I doubt the performance will be that = bad. It depends on the kind of query.

The main problem is= the lack of indexes for literal terms. Since all the "objects" i= n a statement are stored in the same column, we cannot have specialized ind= exes for dates, integers, etc. So when doing a search through for a certain= date range and other similar stuff, it is doing a manual scan.

Also, virtuoso never removes its indexes. It only updates th= em, cause in the "semantic word" deletions are very rare.

=

Cheerio,
Ivan


--
Money can't buy happiness, but neither can poverty.
=C2=A0 -- Leo Rosten

_______________________________________________
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel




--
Vishesh Handa
--001a11c3702c5fa5af04ddc3ec2f-- --===============8062227628940630255== Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Disposition: inline _______________________________________________ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel --===============8062227628940630255==--