Hot data Tracking

Hot data Tracking

[Timo Witte]

[-- Attachment #1: Type: text/plain, Size: 239 bytes --]

What happened to the hot data tracking feature in btrfs? There are a lot
of old patches from aug 2010, but it looks like the feature has been
completly removed from the current version of btrfs. Is this feature
still on the roadmap?


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 900 bytes --]

Re: Hot data Tracking

[David Sterba]

On Sat, Feb 11, 2012 at 05:49:41AM +0100, Timo Witte wrote:
> What happened to the hot data tracking feature in btrfs? There are a lot
> of old patches from aug 2010, but it looks like the feature has been
> completly removed from the current version of btrfs. Is this feature
> still on the roadmap?

Removed? AFAIK it hasn't been ever merged, though it's be a nice
feature. There were suggestions to turn it into a generic API for any
filesystem to use, but this hasn't happened.

The patches are quite independent and it was easy to refresh them on top
of current for-linus branch. A test run did not survive a "random"
xfstest, 013 this time, so I probably mismerged some bits. The patchset
lives in branch foreign/ibm/hotdatatrack in my git repo.


david

Re: Hot data Tracking

[Waxhead]

David Sterba wrote:
> On Sat, Feb 11, 2012 at 05:49:41AM +0100, Timo Witte wrote:
>> What happened to the hot data tracking feature in btrfs? There are a lot
>> of old patches from aug 2010, but it looks like the feature has been
>> completly removed from the current version of btrfs. Is this feature
>> still on the roadmap?
> Removed? AFAIK it hasn't been ever merged, though it's be a nice
> feature. There were suggestions to turn it into a generic API for any
> filesystem to use, but this hasn't happened.
>
> The patches are quite independent and it was easy to refresh them on top
> of current for-linus branch. A test run did not survive a "random"
> xfstest, 013 this time, so I probably mismerged some bits. The patchset
> lives in branch foreign/ibm/hotdatatrack in my git repo.
>
>
> david
>
Someone recently mentioned bcache in another post who seems to cover 
this subject fairly well. However would it not make sense if btrfs 
actually was able to automatically take advantage of whatever disks is 
added to the pool? For example if you have 10 disk of different size and 
performance in a raid5/6 like configuration would it not be feasible if 
btrfs automagically (option) could manage it's own cache? For example it 
could reserve a chunk of free space as cache (based on how much data is 
free) and stripe data over all disks (cache). When the filesystem 
becomes idle or at set intervals it could empty the cache or 
move/rebalance pending writes over to the original raid5/6 like setup.
As far as I remember hot data tracking was all about moving the data 
over to the fastest disk. Why not utilize all disks and benefit from 
disks working together?

Svein Engelsgjerd

Re: Hot data Tracking

[Hubert Kario]

On Thursday 03 of May 2012 15:09:25 Waxhead wrote:
> David Sterba wrote:
> > On Sat, Feb 11, 2012 at 05:49:41AM +0100, Timo Witte wrote:
> >> What happened to the hot data tracking feature in btrfs? There are=
 a lot
> >> of old patches from aug 2010, but it looks like the feature has be=
en
> >> completly removed from the current version of btrfs. Is this featu=
re
> >> still on the roadmap?
> >=20
> > Removed? AFAIK it hasn't been ever merged, though it's be a nice
> > feature. There were suggestions to turn it into a generic API for a=
ny
> > filesystem to use, but this hasn't happened.
> >=20
> > The patches are quite independent and it was easy to refresh them o=
n top
> > of current for-linus branch. A test run did not survive a "random"
> > xfstest, 013 this time, so I probably mismerged some bits. The patc=
hset
> > lives in branch foreign/ibm/hotdatatrack in my git repo.
> >=20
> >=20
> > david
>=20
> Someone recently mentioned bcache in another post who seems to cover
> this subject fairly well.

bcache does one very specific assertion that isn't met by btrfs: overwr=
ing old=20
data in a file writes data to the same place on the disk, same goes for=
=20
metadata. In other words, it won't work with COW file system.

> However would it not make sense if btrfs
> actually was able to automatically take advantage of whatever disks i=
s
> added to the pool? For example if you have 10 disk of different size =
and
> performance in a raid5/6 like configuration would it not be feasible =
if
> btrfs automagically (option) could manage it's own cache? For example=
 it
> could reserve a chunk of free space as cache (based on how much data =
is
> free) and stripe data over all disks (cache). When the filesystem
> becomes idle or at set intervals it could empty the cache or
> move/rebalance pending writes over to the original raid5/6 like setup=
=2E
> As far as I remember hot data tracking was all about moving the data
> over to the fastest disk. Why not utilize all disks and benefit from
> disks working together?

=46or this to work, you need feature set that allows hot data movement =
between=20
disks and data restriping. Then such cache feature will use much of the=
 same=20
code.

Regards,
--=20
Hubert Kario
QBS - Quality Business Software
02-656 Warszawa, ul. Ksawer=F3w 30/85
tel. +48 (22) 646-61-51, 646-74-24
www.qbs.com.pl
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" =
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html