🚀 go-pugleaf

In article <40sv62$nj8@sv01a.cis.squared.com>,
Tony Gast <cis115@ws001a.cis.squared.com> wrote:
> In article <40qs2d$5o9@news.cloud9.net>, tls@cloud9.net (Thor Lancelot Simon) writes:
> |> 
> |> Heavens, no.  Auspex does a perfectly good job of selling their fileservers
> |> themselves.  Just for reference, do you know what an Auspex is?  They're 
> |> specialized network file servers which can cost millions of dollars.  They
> |> also happen to use a correct RAID implementation which does write caching 
> |> without running the risk of losing the whole filesystem.
> |> 
> 
> I recall seeing Auspex advertised.  This sounds even more costly to have
> such a specialized machine, but I am sure there are companies that need
> exactly this sort of dedicated device.

Like us, for example.  What I can't understand from what tls is saying
is why an NVRAM-equipped SCSI controller isn't sufficient to guard
against the scenario he describes (power failure with writes uncommitted
to disk)?  We have a Write Accelerator in our Auspex, and I think I
understand how it works.  Why is an NVRAM+disk combination any less
secure than a disk alone?  Why does the code at the host end of the SCSI
chain need to know that the far end has a cache with NVRAM?  Surely,
provided any acknowledged write down the SCSI chain can be followed
(possibly after a power fail) with a correct read from the same location
on disk, everyone is happy?

Maybe I'm missing some subtle issue, but so far as I am aware an Auspex
write accelerator simply stores uncommitted writes, and only the Storage
Processor (ie not the filesystem code at all) knows about it.  If I
build a box that contains a large NVRAM cache and then does write-behind
into a disk array, how does that break the Unix filesystem code in the
event of a power failure?  Please, I've worked (in the past) on the ufs
filesystem in a port, and I've looked at the bsd filesystem in some
depth: please don't jsut assert, explain.

I assert that the strategy routine will schedule a block write.  That
gets pushed down the SCSI controller.  If there's an NVRAM at the far
end, the write ends up there and the RAID box schedules a disk write for
later.  If the power goes, presumably at power-up the RAID box
autonomously flushes the NVRAM back to disk, which is perfectly safe.  I
don't see the issue.

I'm not interested in buying RAID boxes, as I have a rather wonderful
Auspex.  But I don't see that this issue alone marks ``RAID 7'' as
non-functional.

ian