[Gllug] Partioning advice needed

Nix nix at esperi.org.uk
Mon Feb 19 23:22:02 UTC 2007


On 19 Feb 2007, Mike Brodbelt told this:
> Anthony Newman wrote:
>> 
>> It's odd that init{rd,ramfs} used to be one of those optional things, 
>> now people seem to see it as indispensable whether you actually need it 
>> or not. I just consider it an extra hassle and a hack to make things 
>> work that unfortunately became the norm.
>
> I have always hated initrd, but initramfs is a different story. It
> provides the best of both worlds - you get all your hardware support in
> a single vmlinux file, and you can also get a "recovery console" if
> things go wrong.

You have no *idea* how useful this was when RAIDing my systems for the
first time. (Well, actually, you probably do have an idea because I
think you're using the same scripts. But anyway. :) )

Compared to the boot floppies and huge-delay-knoppix-is-booting-
with-not-quite-the-same-versions-of-anything-as-you-use I'd been putting
up with before, it was heaven.

>> You get what you pay for as usual; the sort of stuff I was alluding to 
>> would probably cost a significant part of a person's lifetime salary 
>> over its 3-5 year lifetime. There's no reason why the technology under 
>> such things won't trickle down into the free implementations though, 
>> provided it isn't too heavily patented.
>
> It mostly has, IMO. Of course a percentage of the cost of the high end
> stuff is raw hardware cost, but there's an obscene amount of markup on
> the stuff.

Part of this huge markup is probably to ensure that all data pending
writing goes into battery-backed RAM until the writing is completed,
because one of the little nasties about RAID-5 is that if you lose power
during a chunk write you can end up with undetected corruption on that
chunk. (Myself I live with the problem: power cuts are rare, and the
same undetected corruption can occur with power cuts while writing on
non-RAIDed systems anyway...)

> Those figures from Ben Rockwood's blog at cuddletech. Half a gig a
> second fast enough? OK, so that's not a cheap piece of hardware

What *bus* are they using? That would max out a PCI several times over.
PCIe?

> else out there. ZFS is one - end to end 64 bit checksumming of all data.
> It's *better* than a parity scrub. But if you want to scrub "zpool scrub
>  filesys" will do it nicely.

Oh yes. Hopefully we'll get something similar on Linux soon (we'd better:
that sort of thing will be *essential* with 2015-capacity disks).

(Thankfully some ex-ZFS hackers, e.g. the inimitable and irrepressible
Val Henson, are working on the problem :) )

> I'm not normally a technical fanboy, and the Solaris userland has a long
> way to go before it could be regarded as pleasant, but there are some
> things that Sun have just done *really* well, and ZFS is IMO the best by
> far.

It looks like dtrace will be superseded in the coolness stakes by
systemtap any time soon (can dtrace trace userspace processes? systemtap
will be able to Real Soon Now), but ZFS, yes, ZFS is one of those things
that has me slobbering. :)

-- 
`In the future, company names will be a 32-character hex string.'
  --- Bruce Schneier on the shortage of company names
-------------- next part --------------
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug


More information about the GLLUG mailing list