--===============1022713810425747842==
Content-Type: multipart/signed; boundary="nextPart3471080.GSDXTeT1KQ"; micalg="pgp-sha1"; protocol="application/pgp-signature"


--nextPart3471080.GSDXTeT1KQ
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="iso-8859-1"

On Wednesday 26 of March 2014 13:06:23 Martin Steigerwald wrote:
> Hi!
>=20
> I reported
>=20
> Bug 332626 - MySQL tuning: adaption of MySQL tuning options for large=
r
> accounts
> https://bugs.kde.org/show_bug.cgi?id=3D332626
>=20
>
> But its somewhat arguable, I know.
>=20

Hi,

I actually talked about optimizing mysql.conf with MariaDB people on FO=
SDEM in=20
February.

> I think I saw quite an improvement with raising the default low innod=
b
> buffer pool size of Akonadi of 80 MiB, which is even below the MySQL
> default of 128 MiB, to 256 MiB helped with a 11 GiB large POP3 accoun=
t,
> especially when synchronizing linux-kernel-ml folder with about 22300=
0
> unread mails. And also on retrieving and filtering new mails there, w=
hich
> can make KMail quite unresponsive on a quite powerful machine (Sandyb=
ridge,
> 8 GB RAM, dual SSD BTRFS RAID 1). I will do some testing, yet it is
> difficult to know for sure. mysqltuner.pl script gave some clear
> recommendations tough, which are to high for a desktop machine tough,=
 I
> think (>1 GiB of innodb buffer pool size). It seems it recommends adj=
usting
> to actual data size in database, which would only be suitable if all =
of the
> data is accessed often.

The problem with innodb_buffer_pool_size is, that it directly impacts t=
he=20
amount of RAM the mysqld process takes, which  makes it difficult to fi=
nd a =20
good compromise between what LKML subscribers need and what users with =
couple=20
of calendars and an adressbook need.

> What is the amount data Akonadi accesses frequently? Are there any
> experiences with that? I bet it depends highly on how the user uses i=
t and
> the amount of mails in a single folder to be synchronized.

This is difficult to say and I don't think anyone ever tried to benchma=
rk or=20
measure it. Indeed it does depend on how you use Akonadi, an

When you open a folder in KMail, we run

SELECT * FROM PartTable WHERE parentCollection =3D XY AND partType =3D=20=

"PLD:ENVELOPE"

which will return all envelopes in the collection (that's subject, date=
, from,=20
to, cc, bcc, ...)

akonadi=3D# SELECT MIN(datasize), MAX(datasize), AVG(datasize) FROM Par=
tTable=20
WHERE partTypeId =3D 3;
 min |  max  |         avg         =20
=2D----+-------+----------------------
  45 | 23354 | 487.8883434774372617
(1 row)

akonadi=3D# SELECT datasize AS median FROM PartTable WHERE partTypeId =3D=
3 ORDER=20
BY datasize DESC OFFSET (SELECT (COUNT(*) + 1) / 2 FROM PartTable WHERE=
=20
partTypeId =3D 3) LIMIT 1;
 median=20
=2D-------
    451

(partTypeId =3D 3 =3D=3D SELECT id FROM PartTypeTable WHERE ns =3D 'PLD=
' AND=20
name=3D'ENVELOPE')

Which makes roughly 6.5MB of data in a collection of 15 000 items. I do=
n't=20
think we have any application (except for the indexer) that would reque=
st the=20
full payload of all items in a collection of such size. I know KAddress=
book=20
and KOrganizer do that, but the collections there are much smaller, and=
 so are=20
the actual payloads.

>=20
> What do you think?

I think we can try to play with the values a bit, having some kind of=20=

benchmark or test tool would be nice :-)

> Do you have any plans to switch to MariaDB and probably investigating=

> storage engines which may be more suitable for tuning to desktop load=
s?

MariaDB is a drop-in replacement for MySQL, so it's supported by defaul=
t (I've=20
been using MariaDB instead of MySQL for over year before I switched to =
PSQL)

AFAIK the only engine that we can really use is XtraDB, which is just a=
n=20
enhanced fork of InnoDb.

> InnoDB is difficult as innodb_buffer_pool_size needs to take free mem=
ory
> into account which can change quite rapidly on desktops or anything e=
lse
> than a dedicated database server. A engine which uses Linux pagecache=
 for
> most of its caching would be interesting I think.

Maybe we could have some kind of "initial setup" tool that would check=20=

available memory and would update innodb_buffer_pool_size depending on =
that=20
(taking in account some max limit, min limit, max percentage, ...)


Cheers,
Dan

>=20
>=20
> Just wanted to raise this topic.
>
> Ciao,
> Martin

=2D-=20
Daniel Vr=E1til | dvratil@redhat.com | dvratil on #kde-devel, #kontact,=
 #akonadi
KDE Desktop Team
Associate Software Engineer, Red Hat, Inc.

GPG Key: 0xC59D614F6F4AE348
Fingerprint: 4EC1 86E3 C54E 0B39 5FDD B5FB C59D 614F 6F4A E348
--nextPart3471080.GSDXTeT1KQ
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: This is a digitally signed message part.
Content-Transfer-Encoding: 7Bit

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)

iQEcBAABAgAGBQJTMt5tAAoJEMWdYU9vSuNInY0IAIEczGkElalOAd10de3AejGH
WAKqajpBDCUgrVVLPiO0FLLBwX6KUKvMADFWHNbR1giW8t9maWV8tqLPi3i6QGD9
EesOSfLpdeaM00i1T53Mv/zXSzek3cs5GQQayRIJWbzljAgI2B/UdZL4WmI0xsFS
IyNl6ElM9S+utKXRZkMgq1G3PDWJwQndTNG/198vQidWRxqPUt5uVUukp5GgUzJM
9u5u4h43/5qhD1yrINFZRdt+VGMDPqQf5kzxqhvb9xTUQ3Hr1v194MfSYQdQxqR4
rMdAs6Ci8k70tuAo3YYSCwTAVxDTHhm1eIO0xS2GcsnE3RQxKsvvsmlojKoJqmA=
=obSG
-----END PGP SIGNATURE-----

--nextPart3471080.GSDXTeT1KQ--


--===============1022713810425747842==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
KDE PIM mailing list kde-pim@kde.org
https://mail.kde.org/mailman/listinfo/kde-pim
KDE PIM home page at http://pim.kde.org/
--===============1022713810425747842==--