write-intent bitmaps

write-intent bitmaps

[Russell Coker]

http://lists.debian.org/debian-devel/2008/01/msg00921.html

Are they regarded as a stable feature?  If so I'd like to see distributions 
supporting them by default.  I've started a discussion in Debian on this 
topic, see the above URL for details.

-- 
russell@coker.com.au
http://etbe.coker.com.au/          My Blog

http://www.coker.com.au/sponsorship.html Sponsoring Free Software development

Re: write-intent bitmaps

[Neil Brown]

On Sunday January 27, russell@coker.com.au wrote:
> http://lists.debian.org/debian-devel/2008/01/msg00921.html
> 
> Are they regarded as a stable feature?  If so I'd like to see distributions 
> supporting them by default.  I've started a discussion in Debian on this 
> topic, see the above URL for details.

Yes, it is regarded as stable.

However it can be expected to reduce write throughput.  A reduction of
several percent would not be surprising, and depending in workload it
could probably be much higher.

It is quite easy to add or remove a bitmap on an active array, so
making it a default would probably be fine providing it was easy for
an admin to find out about it and remove the bitmap is they wanted the
extra performance.

NeilBrown

Re: write-intent bitmaps

[Russell Coker]

On Sunday 27 January 2008 22:21, Neil Brown <neilb@suse.de> wrote:
> On Sunday January 27, russell@coker.com.au wrote:
> > http://lists.debian.org/debian-devel/2008/01/msg00921.html
> >
> > Are they regarded as a stable feature?  If so I'd like to see
> > distributions supporting them by default.  I've started a discussion in
> > Debian on this topic, see the above URL for details.
>
> Yes, it is regarded as stable.

Thanks for that information.

> However it can be expected to reduce write throughput.  A reduction of
> several percent would not be surprising, and depending in workload it
> could probably be much higher.

It seems to me that losing a few percent of performance all the time is better 
than a dramatic performance loss for an hour or two when things go wrong.

> It is quite easy to add or remove a bitmap on an active array, so
> making it a default would probably be fine providing it was easy for
> an admin to find out about it and remove the bitmap is they wanted the
> extra performance.

I hadn't realised that.  So having this in the installer is not as important 
as I previously thought.

-- 
russell@coker.com.au
http://etbe.coker.com.au/          My Blog

http://www.coker.com.au/sponsorship.html Sponsoring Free Software development

Re: write-intent bitmaps

[Russell Coker]

On Tuesday 29 January 2008 05:15, Bill Davidsen <davidsen@tmr.com> wrote:
> You may have missed the "much higher" part of the previous paragraph.
> And given the reliability of modern drives, unless you have a LOT of
> them you may be looking at years of degraded performance to save a few
> hours of slow performance after a power fail or similar. In other words,
> it's not as black and white as it seems.

What is the pathological case?  1/2 or 1/3 write performance?

For serious write performance of a RAID you want a NVRAM write-back cache for 
RAID-5 stripes, and the NVRAM cache removes the need for write-intent 
bitmaps.  AFAIK Linux software RAID doesn't support such things and that 
putting filesystem journals and the write-intent bitmap blocks on NVRAM 
devices is the best that you could achieve.

It seems that if you want the best performance for small synchronous writes 
(EG a mail server - which may be the most pessimal application for 
write-intent bitmaps) then hardware RAID is the only option.

Are there plans for supporting a NVRAM write-back cache with Linux software 
RAID?

-- 
russell@coker.com.au
http://etbe.coker.com.au/          My Blog

http://www.coker.com.au/sponsorship.html Sponsoring Free Software development

Re: write-intent bitmaps

[Peter Rabbitson]

Russell Coker wrote:
> Are there plans for supporting a NVRAM write-back cache with Linux software 
> RAID?
> 

AFAIK even today you can place the bitmap in an external file residing on a 
file system which in turn can reside on the nvram...

Peter

Re: write-intent bitmaps

[Russell Coker]

On Tuesday 29 January 2008 20:13, Peter Rabbitson <rabbit+list@rabbit.us> 
wrote:
> Russell Coker wrote:
> > Are there plans for supporting a NVRAM write-back cache with Linux
> > software RAID?
>
> AFAIK even today you can place the bitmap in an external file residing on a
> file system which in turn can reside on the nvram...

True, and you can also put the journal of a filesystem on a NVRAM device.  But 
that doesn't give the stripe aggregating benefits for RAID-5 or the general 
write-back cache benefits for everything else.

-- 
russell@coker.com.au
http://etbe.coker.com.au/          My Blog

http://www.coker.com.au/sponsorship.html Sponsoring Free Software development