Block layout rpm

Block layout rpm

[Jim Rees]

I have released a developer's snapshot Fedora rpm of nfs-utils for the pnfs
block layout client, at:
http://www.citi.umich.edu/projects/nfsv4/pnfs/block/

Instructions, as always, are on the wiki:
http://wiki.linux-nfs.org/wiki/index.php/Fedora_pNFS_Client_Setup

While this rpm does contain fixes that aren't yet in Steve's Fedora pnfs
rpm, it is really only intended for block layout testing.  If you do try it,
please report results (positive or negative) to me.

Re: Block layout rpm

[Jeff Garzik]

On 11/29/2010 04:07 PM, Jim Rees wrote:
> I have released a developer's snapshot Fedora rpm of nfs-utils for the pnfs
> block layout client, at:
> http://www.citi.umich.edu/projects/nfsv4/pnfs/block/
>
> Instructions, as always, are on the wiki:
> http://wiki.linux-nfs.org/wiki/index.php/Fedora_pNFS_Client_Setup

"pNFS will work with 32-bit linux and lesser hardware, but the you will 
not obtain the performance benefits of pNFS."

That's just silly, and, I might even go so far as to say "wrong".  The 
benefits of direct-to-storage access, without an MDS middle-man, are 
quite noticable even for 32-bit, underpowered, and/or embedded clients.

	Jeff

Re: Block layout rpm

[Jim Rees]

Jeff Garzik wrote:

  "pNFS will work with 32-bit linux and lesser hardware, but the you
  will not obtain the performance benefits of pNFS."
  
  That's just silly, and, I might even go so far as to say "wrong".
  The benefits of direct-to-storage access, without an MDS middle-man,
  are quite noticable even for 32-bit, underpowered, and/or embedded
  clients.

Agreed.  It used to say 64-bit only, and I added the part about 32-bit
because it does run on 32-bit, but really that should just come out.  I'll
fix it.

Re: Block layout rpm

[Jeff Garzik]

On 11/29/2010 05:26 PM, Jim Rees wrote:
> Jeff Garzik wrote:
>
>    "pNFS will work with 32-bit linux and lesser hardware, but the you
>    will not obtain the performance benefits of pNFS."
>
>    That's just silly, and, I might even go so far as to say "wrong".
>    The benefits of direct-to-storage access, without an MDS middle-man,
>    are quite noticable even for 32-bit, underpowered, and/or embedded
>    clients.
>
> Agreed.  It used to say 64-bit only, and I added the part about 32-bit
> because it does run on 32-bit, but really that should just come out.  I'll
> fix it.

Thanks.

I'd also say, with my networking hat on, the VLAN statement is 
unqualified.  VLAN is not necessary for pNFS, but the way the statement 
reads to me, I get the opposite impression.

pNFS works quite well if you're just a 1U client with a single NIC, for 
example.

	Jeff