pnetcdf and pNFS?

[Michalakes, John]

Thanks Rob. I don't think we rely on any read/write access, but I'll have to ponder.  Assuming that's the case and that we're doing only read-only or write-only to a given file, have you found that pNetCDF on pNFS is reasonably fast and scalable (relative to other parallel file systems)?  Thanks, -John

-----Original Message-----
From: Rob Ross [mailto:rross at mcs.anl.gov] 
Sent: Monday, July 26, 2010 3:27 PM
To: Michalakes, John
Cc: parallel-netcdf at mcs.anl.gov
Subject: Re: pnetcdf and pNFS?

The short answer is that I think it should be ok if you are doing read- 
only or write-only access.

pNFS is NFSv4, which by default has close-to-open semantics as I  
understand it. However, the NFSv4 standard writers were careful to  
make sure that clients would not inadvertently touch bytes they  
shouldn't when writing (i.e., not touch bytes that weren't written via  
R/M/W).

close-to-open semantics mean that clients aren't guaranteed to see  
changes to the file that occur between an open and a close. Thus read/ 
write access results are probably undefined.

If someone knows better, I'm happy to be corrected!

Rob

On Jul 26, 2010, at 12:43 PM, Michalakes, John wrote:

> Hi,
>
> Wondering if you have any information on suitability of pNFS for use  
> with pNetCDF?
>
> Thanks,
>
> John
>
> John Michalakes
> National Renewable Energy Laboratory
> 1617 Cole Blvd.
> Golden, Colorado 80401
> Phone: 303-275-4297
> Fax: 303-275-4091
> John.Michalakes at nrel.gov
>