On Mon, 14 Sep 1998, Matt Thomas wrote:
>>  It's a chunk of battery-backed RAM that's used as a write cache for NFS data. 
>>Parts of the NFS code are modified to write to this cache and then return from a
>>write call immediately, instead of waiting for it to get all the way to disk. 
>>The battery-backed RAM is flushed periodically.
>
>Not quite right.  The buffer cache/VFS layer is modified to redirect synchronous
>data writes to the Prestoserve NVRAM and schedule async writes from the NVRAM.
>Hence NFS does not even know it's there.  This also helped other software which
>caused a lot of sync writes to happen (like directory updates on news servers).

  Huh?  Hmm.  Under SunOS (the only OS under which I've run a Prestoserve) nfsd
is completely replaced.

>>  If the machine goes down, stuff in the rc scripts checks for stale data in
>>that chunk of RAM...if there is some, it flushes it out to where it's supposed
>>to be.
>
>Actually, fsck took care of it under ULTRIX.  It would apply the changes from
>Prestoserve NVRAM before actually trying to verify the disk structure.

  Ahh, weird.  The last time I used one (again under SunOS, which is probably
the difference) there was a program that ran out of one fo the rc scripts that
did the check & flush...

                              -Dave McGuire