Hi Reiserfs-dev

Hi Reiserfs-dev

[Steven]

Been using Reiser 3 for years now, messing round with m-multiple operating systems. Imho, it's easily the best general purpose linux filesystem. Does anyone here say otherwise ?

I went through the hassle of installing reiser 4 on 2.6.24 about a year ago, but it seemed to have latency issues, esp. at start-up (fedora 4, 1.13 PIII dell laptop with 256ram) and reverted back to reiser 3. Ideas ?

Thanks, Steven.


      

Re: Hi Reiserfs-dev

[Zan Lynx]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Steven wrote:
> Been using Reiser 3 for years now, messing round with m-multiple 
> operating systems. Imho, it's easily the best general purpose linux 
> filesystem.  Does anyone here say otherwise ?
> 
> I went through the hassle of installing reiser 4 on 2.6.24 about a 
> year ago, but it seemed to have latency issues, esp. at start-up 
> (fedora 4, 1.13 PIII dell laptop with 256ram) and reverted back to 
> reiser 3. Ideas?
> 
> Thanks, Steven.

I've been running Reiser4 on my laptop as / for a long time now. I have
two bits of advice.

Make sure it mounts at boot with the noatime option. Something like
"rootflags=noatime" on your LILO or GRUB kernel options line.

Have your init scripts set the queue depth on your block IO scheduler to
1024. Something like "echo 1024 > /sys/block/sda/queue/nr_requests".
With just 256 MB you might want to try smaller values. I'm not very
clear on how much RAM each queue slot can use.

I found that Reiser4 will appear to hang if it cannot flush an entire
transaction atom to the IO queue in one go. Make sure it's big enough.

The other thing that seems to make it a little slow is that it uses
barriers.  This is a Linux feature that serious filesystems use to
guarantee data integrity.  ext3 uses barriers for its journal and XFS
uses barriers too.  I believe there were patches for reiser3 to use
barriers too but I don't know if those are in all Linux or just SUSE.

As I understand it, on an IDE drive, a barrier is pretty slow because it
forces a full write queue flush, does the write, then flushes the queue
*again* before continuing.  On a real SCSI device with tagged queuing
this works much more quickly because it can continue to issue disk
commands after the barrier.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAki0o8gACgkQolqWs/Y4NLy9MgCgh3v5mB4uLx673JT1sz6dgJyA
1XEAni0AZxC6DokBQQ+sDu18qbzxP0s5
=K2Cy
-----END PGP SIGNATURE-----

Re: Hi Reiserfs-dev

[Volker Armin Hemmann]

On Mittwoch, 27. August 2008, Zan Lynx wrote:
> Steven wrote:
> > Been using Reiser 3 for years now, messing round with m-multiple
> > operating systems. Imho, it's easily the best general purpose linux
> > filesystem.  Does anyone here say otherwise ?
> >
> > I went through the hassle of installing reiser 4 on 2.6.24 about a
> > year ago, but it seemed to have latency issues, esp. at start-up
> > (fedora 4, 1.13 PIII dell laptop with 256ram) and reverted back to
> > reiser 3. Ideas?
> >
> > Thanks, Steven.
>
> I've been running Reiser4 on my laptop as / for a long time now. I have
> two bits of advice.
>
> Make sure it mounts at boot with the noatime option. Something like
> "rootflags=noatime" on your LILO or GRUB kernel options line.
>
> Have your init scripts set the queue depth on your block IO scheduler to
> 1024. Something like "echo 1024 > /sys/block/sda/queue/nr_requests".
> With just 256 MB you might want to try smaller values. I'm not very
> clear on how much RAM each queue slot can use.
>
> I found that Reiser4 will appear to hang if it cannot flush an entire
> transaction atom to the IO queue in one go. Make sure it's big enough.
>
> The other thing that seems to make it a little slow is that it uses
> barriers.  This is a Linux feature that serious filesystems use to
> guarantee data integrity.  ext3 uses barriers for its journal and XFS
> uses barriers too.  I believe there were patches for reiser3 to use
> barriers too but I don't know if those are in all Linux or just SUSE.

reiserfs uses barriers for ages. Xfs and reiserfs use barriers by default.
ext3 does not use barriers by default. Because of the performance hit.

Every 'benchmark' that does not enable or disable barriers is skewed in ext3's 
advantage. With barriers turned on for all filesystems ext3 is very slow.

Re: Hi Reiserfs-dev

[Steven]

Thanks for the replies :) Thats the first time i've heard about barriers.

I've always used noatime.. so thats no the issue. (I recently learnt to disable the dodgy Fedora readahead service too, and can't remember if i knew about this back then). Part of the problem is Reiser3 is so damn good (speed and robustness), and even FreeBSD has RO support nowadays, it's going to be hard to change even though reiser4's tail packing always got me excited.

--- On Wed, 8/27/08, Volker Armin Hemmann <volker.armin.hemmann@tu-clausthal.de> wrote:

> From: Volker Armin Hemmann <volker.armin.hemmann@tu-clausthal.de>
> Subject: Re: Hi Reiserfs-dev
> To: reiserfs-devel@vger.kernel.org
> Date: Wednesday, August 27, 2008, 2:48 PM
> On Mittwoch, 27. August 2008, Zan Lynx wrote:
> > Steven wrote:
> > > Been using Reiser 3 for years now, messing round
> with m-multiple
> > > operating systems. Imho, it's easily the best
> general purpose linux
> > > filesystem.  Does anyone here say otherwise ?
> > >
> > > I went through the hassle of installing reiser 4
> on 2.6.24 about a
> > > year ago, but it seemed to have latency issues,
> esp. at start-up
> > > (fedora 4, 1.13 PIII dell laptop with 256ram) and
> reverted back to
> > > reiser 3. Ideas?
> > >
> > > Thanks, Steven.
> >
> > I've been running Reiser4 on my laptop as / for a
> long time now. I have
> > two bits of advice.
> >
> > Make sure it mounts at boot with the noatime option.
> Something like
> > "rootflags=noatime" on your LILO or GRUB
> kernel options line.
> >
> > Have your init scripts set the queue depth on your
> block IO scheduler to
> > 1024. Something like "echo 1024 >
> /sys/block/sda/queue/nr_requests".
> > With just 256 MB you might want to try smaller values.
> I'm not very
> > clear on how much RAM each queue slot can use.
> >
> > I found that Reiser4 will appear to hang if it cannot
> flush an entire
> > transaction atom to the IO queue in one go. Make sure
> it's big enough.
> >
> > The other thing that seems to make it a little slow is
> that it uses
> > barriers.  This is a Linux feature that serious
> filesystems use to
> > guarantee data integrity.  ext3 uses barriers for its
> journal and XFS
> > uses barriers too.  I believe there were patches for
> reiser3 to use
> > barriers too but I don't know if those are in all
> Linux or just SUSE.
> 
> reiserfs uses barriers for ages. Xfs and reiserfs use
> barriers by default.
> ext3 does not use barriers by default. Because of the
> performance hit.
> 
> Every 'benchmark' that does not enable or disable
> barriers is skewed in ext3's 
> advantage. With barriers turned on for all filesystems ext3
> is very slow.
> 
> 
> 
> --
> To unsubscribe from this list: send the line
> "unsubscribe reiserfs-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at 
> http://vger.kernel.org/majordomo-info.html


      

Re: Hi Reiserfs-dev

[Edward Shishkin]

Steven wrote:
> Thanks for the replies :) Thats the first time i've heard about barriers.
>
> I've always used noatime.. so thats no the issue. (I recently learnt to disable the dodgy Fedora readahead service too, and can't remember if i knew about this back then). Part of the problem is Reiser3 is so damn good (speed and robustness), and even FreeBSD has RO support nowadays, it's going to be hard to change even though reiser4's tail packing always got me excited.
>
>   

Actually, tail _conversion_ (not packing).

Working around dirty polymorphisms in the ITEM interface (a file built
of tails can not be mmap-ed with flags VM_MAYWRITE | VM_SHARED)
plus working around deadlock issues make its implementation very
complex.

So I would prefer reiser4 to go to mainline without tail stuff. However it
will require to fork the stuff for people who are not ready to reformat
their partitions, and it is not good, because somebody should take
care of the second branch. Any ideas?

Edward.

> --- On Wed, 8/27/08, Volker Armin Hemmann <volker.armin.hemmann@tu-clausthal.de> wrote:
>
>   
>> From: Volker Armin Hemmann <volker.armin.hemmann@tu-clausthal.de>
>> Subject: Re: Hi Reiserfs-dev
>> To: reiserfs-devel@vger.kernel.org
>> Date: Wednesday, August 27, 2008, 2:48 PM
>> On Mittwoch, 27. August 2008, Zan Lynx wrote:
>>     
>>> Steven wrote:
>>>       
>>>> Been using Reiser 3 for years now, messing round
>>>>         
>> with m-multiple
>>     
>>>> operating systems. Imho, it's easily the best
>>>>         
>> general purpose linux
>>     
>>>> filesystem.  Does anyone here say otherwise ?
>>>>
>>>> I went through the hassle of installing reiser 4
>>>>         
>> on 2.6.24 about a
>>     
>>>> year ago, but it seemed to have latency issues,
>>>>         
>> esp. at start-up
>>     
>>>> (fedora 4, 1.13 PIII dell laptop with 256ram) and
>>>>         
>> reverted back to
>>     
>>>> reiser 3. Ideas?
>>>>
>>>> Thanks, Steven.
>>>>         
>>> I've been running Reiser4 on my laptop as / for a
>>>       
>> long time now. I have
>>     
>>> two bits of advice.
>>>
>>> Make sure it mounts at boot with the noatime option.
>>>       
>> Something like
>>     
>>> "rootflags=noatime" on your LILO or GRUB
>>>       
>> kernel options line.
>>     
>>> Have your init scripts set the queue depth on your
>>>       
>> block IO scheduler to
>>     
>>> 1024. Something like "echo 1024 >
>>>       
>> /sys/block/sda/queue/nr_requests".
>>     
>>> With just 256 MB you might want to try smaller values.
>>>       
>> I'm not very
>>     
>>> clear on how much RAM each queue slot can use.
>>>
>>> I found that Reiser4 will appear to hang if it cannot
>>>       
>> flush an entire
>>     
>>> transaction atom to the IO queue in one go. Make sure
>>>       
>> it's big enough.
>>     
>>> The other thing that seems to make it a little slow is
>>>       
>> that it uses
>>     
>>> barriers.  This is a Linux feature that serious
>>>       
>> filesystems use to
>>     
>>> guarantee data integrity.  ext3 uses barriers for its
>>>       
>> journal and XFS
>>     
>>> uses barriers too.  I believe there were patches for
>>>       
>> reiser3 to use
>>     
>>> barriers too but I don't know if those are in all
>>>       
>> Linux or just SUSE.
>>
>> reiserfs uses barriers for ages. Xfs and reiserfs use
>> barriers by default.
>> ext3 does not use barriers by default. Because of the
>> performance hit.
>>
>> Every 'benchmark' that does not enable or disable
>> barriers is skewed in ext3's 
>> advantage. With barriers turned on for all filesystems ext3
>> is very slow.
>>
>>
>>
>> --
>> To unsubscribe from this list: send the line
>> "unsubscribe reiserfs-devel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at 
>> http://vger.kernel.org/majordomo-info.html
>>     
>
>
>       
> --
> To unsubscribe from this list: send the line "unsubscribe reiserfs-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>   

Re: Hi Reiserfs-dev

[Steven]

> > even though reiser4's tail packing always got me excited.
> Actually, tail _conversion_ (not packing).

Is this correct ? - R3 has tail packing , and tail coversion is
one of the major new features in R4. What is "tail
conversion" ?

> As I understand it, on an IDE drive, a barrier is pretty
> slow because it forces a full write queue flush, does the
> write, then flushes the queue *again* before continuing.
> On a real SCSI device with tagged queuing this works much
> more quickly because it can continue to issue disk

Looking briefly at the 2.6.24 doco and my dmesgs, it's not
obvious if barriers are used by default.  Is it be advisable
to mess around with barriers=0 on a sata drive (/dev/sda).
Are sata drivers totally different to ide and real scsi ?

Steven

------------------------

[ I can't find much stats in /proc. For reference, here's dmesg:

scsi0 : sata_via
scsi1 : sata_via
ata1: SATA max UDMA/133 cmd 0x9800 ctl 0x9c00 bmdma 0xa800 irq 11
ata2: SATA max UDMA/133 cmd 0xa000 ctl 0xa400 bmdma 0xa808 irq 11
ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
ata1.00: HPA detected: current 976771055, native 976773168
ata1.00: ATA-7: SAMSUNG HD502IJ, 1AA01109, max UDMA7
ata1.00: 976771055 sectors, multi 16: LBA48 NCQ (depth 0/32)
ata1.00: configured for UDMA/133
ata2: SATA link down 1.5 Gbps (SStatus 0 SControl 300)
scsi 0:0:0:0: Direct-Access     ATA      SAMSUNG HD502IJ
1AA0 PQ: 0 ANSI: 5
sd 0:0:0:0: [sda] 976771055 512-byte hardware sectors (500107 MB)
sd 0:0:0:0: [sda] Write Protect is off
sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00
sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
sd 0:0:0:0: [sda] 976771055 512-byte hardware sectors (500107 MB)
sd 0:0:0:0: [sda] Write Protect is off
sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00
sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
sda: sda1 sda2 sda3 < sda5 sda6 sda7 sda8 > sda4
sda4: <bsd: sda9 sda10 >
sd 0:0:0:0: [sda] Attached SCSI disk
sd 0:0:0:0: Attached scsi generic sg0 type 0

My mobo is a k8vm800m (VIA K8M800 Chipset. VIA VT8237 Chipset).  ]