[Kde-pim] Akonadi: single database design mistake?

[Martin Steigerwald]

Am Montag, 29. April 2013, 16:37:41 schrieb Milian Wolff:
> On Monday 29 April 2013 07:30:27 bhlevca wrote:
> > Andras Mantia-2 wrote
> >
> > 
> >
> > > Point out the design problem and than we can talk about something,
> > > without
> > > it, sorry, this not more, but the usual ranting without supporting
> > > evidence
> > > of what is the problem...
> >
> > 
> >
> > Why did I expect a decent answer? I don't know, but this is just another
> > rude, mindless email answer defending a bad decision and refusing to see
> > that everyone is complaining about KDEPIM.
> >
> > 
> >
> > If you want a supporting evidence like code lines, you know very well that
> > it is very hard to give and would take time. I would rather rewrite the
> > whole thing if I had the time.
> >
> > 
> >
> > But,  If you take the time and cool down as I initially suggested, you
> > will
> > notice that in my first message I suggested alternatives (the notmuch mail
> > approach) and I pinpointed the culprit: The large behemoth central
> > database
> > based design that does everything and is the bottleneck. And I repeat,
> > although it is a nice idea, its complexity will bring you down if it
> > didn't
> > already.
> 
> Considering that I did a bunch of profile runs on kdepim, I rarely, if
> ever,  saw a bottleneck in the database. Mysqld is also running just fine.

I beg to differ here. And I think I can prove it (via atop logs):

[Akonadi] [Bug 319208] New: KMail unresponsive for minutes after POP3 
retrieval and filtering with high MySQL load for minutes

https://bugs.kde.org/show_bug.cgi?id=319208

(This is with full text mail indexing *disabled*)


And yes, I have the impression that raising innodb_buffer_pool_size from an 
IMHO way to low 80M for decently sized mail accounts to 500M, seemed to help.

Its still to early to say for sure, but it seems to be more responsive. In the 
next days, especially on retrieving POP3 mails accumulated over the night I 
can say more.

I will try for some days and probably reduce it in steps by 100 MiB as long as 
performance is still okay for this speed machine (see below).


Still, I think it is not a design problem. But I do think, that there are 
still database configuration and implementation issues.

This is on a ThinkPad T520 with 8 GiB and 300 GB Intel SSD 320 that might 
reach a similar performance than the VM with out Zimbra server at work which 
serves about 50 employees and also serves mail folders with >350000 mails just 
in time. Well the VM with the NetApp FAS storage behind it, might still be 
faster, but then this SSD takes quite something.

So or so: The hardware is capable enough for handling 6,7 GB of mails. KMail 
did it. If KMail is unresponsive for minutes, its a bug in the software.

Thats my oppinion, and I think its important to see that there are still 
issues with performance. I also do have some more bug reports open it.

I think there are also still issues with correctness. I opened about 20 bug 
reports since my initial tries with KDEPIM-2 and I think that means something.

Thanks,
-- 
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7
_______________________________________________
KDE PIM mailing list kde-pim at kde.org
https://mail.kde.org/mailman/listinfo/kde-pim
KDE PIM home page at http://pim.kde.org/