Re: [nfsv4] IETF 78 NFSv4 session currently schedule for Wednesday afternoon - 1-3:30pm

On 7/6/2010 10:01 AM, Noveck_David@emc.com wrote:
> Making progress on v4.2 sounds sensible but I don't see allocating a large portion of 85 minutes to what might turn out to be an unstructured discussion of possible additions to NFSv4.2 (assuming that we are now OK as far as this being on the working group's charter).  We could certainly use up the time but I don't see that more than 20 minutes would be useful in trying to come up with a proposed "final" list for group discussion.
>
> If people have proposed additions, I should think that they would be able to put together a few slides and propose it as part of the meeting, before this 20-minute session.  I think that that session should be filtering things already proposed and not accumulating last-minute proposals.
>    

On this note, I actually would like to request 5-10 minutes to discuss 
another modest proposal (in addition to our existing proposal to handle 
sparse file located here 
http://tools.ietf.org/html/draft-hildebrand-nfsv4-read-sparse-00)

I would like to propose a method for NFS to perform direct I/O on the 
server to the exported file system.
Direct I/O (O_DIRECT) is a common way for applications that either 
manage caching themselves, perform
large I/O, or realize poor cache hit rates to avoid the overhead of 
using the page cache.  Example applications
include HPC, DB, and virtual machines.  In NFS, when applications open a 
file with the O_DIRECT flag, the
NFS client bypasses the page cache and delivers data directly to/from 
the application.  But in this scenario,
the server file system is unaware of the client's behaviour and 
continues to cache data blocks, possibly
reducing I/O performance and/or polluting the server cache.

We propose to modify either the OPEN RPC (or both the READ and WRITE 
RPCs) to pass the
O_DIRECT flag from the NFS client to the NFS server as a hint.  The NFS 
server can then
pass this flag to the underlying file system when it opens the file.  
This gives the underlying server
file system the option to either perform or not perform direct I/O on 
all I/O to this file.

Dean Hildebrand
IBM Almaden

>
> -----Original Message-----
> From: nfsv4-bounces@ietf.org [mailto:nfsv4-bounces@ietf.org] On Behalf Of Dean Hildebrand
> Sent: Tuesday, July 06, 2010 12:40 PM
> To: nfsv4@ietf.org
> Subject: Re: [nfsv4] IETF 78 NFSv4 session currently schedule for Wednesday afternoon - 1-3:30pm
>
> On 7/4/2010 1:41 PM, Sorin Faibish wrote:
>    
>> On Sun, 04 Jul 2010 16:29:44 -0400, Tom Haynes<tom.haynes@oracle.com>
>> wrote:
>>
>>      
>>> That is close to 1/2 of the meeting.
>>>        
>> Not sure I follow: 15+20+30=65 out of 150. Are there other agenda
>> topics that will take
>> more than 85 minutes? If so we can make room for all the topics.
>>      
> I'll be in Maastricht, and I'm hoping that a good portion of the meeting
> will be deciding on the final items to be included in NFSv4.2.  My
> understanding is that most of the known items have been proposed and
> presented at previous meetings, and so this should be a good opportunity
> to present any remaining items and propose a final list.
>
> Dean
>
>    
>>      
>>> We either need more time or need to frontload the discussion online.
>>>
>>> We had talked about both options at IETF 78...
>>>        
>> I just presented the permission draft, now after the discussion in the
>> email we came to a final
>> proposal before last call which we want to discuss. As about the
>> storage preferences we have
>> to discuss the concrete options we are proposing to discussion before
>> the draft. As you can
>> remember we had a lot of corridor discussions that we need to have in
>> the WG. We can reduce the
>> time if there are conflicts and time shortage.
>>
>> /Sorin
>>
>>      
>>> On Jul 4, 2010, at 8:22 AM, Sorin Faibish<sfaibish@emc.com>  wrote:
>>>
>>>        
>>>> Now that Trond opened this I would like to have 20 min to talk about
>>>> the permission access draft
>>>> that we intent to finalyze after this meeting. We (me and David
>>>> Black) want to have 30min
>>>> discussion on the Storage Preferences for virtualization
>>>> (continuation of the last
>>>> discussion from Anaheim) but in the room not outside like in
>>>> Anaheim. David Black will
>>>> present few slides to open the discussion.
>>>>
>>>> Thanks Spencer and congrats on your new position.
>>>>
>>>> /Sorin
>>>>
>>>> On Sat, 03 Jul 2010 18:01:21 -0400, Trond Myklebust
>>>> <trond.myklebust@fys.uio.no>  wrote:
>>>>
>>>>          
>>>>> On Thu, 2010-06-24 at 18:10 -0500, Spencer wrote:
>>>>>            
>>>>>> Full draft agenda may be found here:
>>>>>> https://datatracker.ietf.org/meeting/78/agenda.html
>>>>>>
>>>>>> _______________________________________________
>>>>>> nfsv4 mailing list
>>>>>> nfsv4@ietf.org
>>>>>> https://www.ietf.org/mailman/listinfo/nfsv4
>>>>>>              
>>>>> Hi Spencer,
>>>>>
>>>>> I'd like a 10-15 minute slot to talk about some NFSv4 migration
>>>>> issues.
>>>>>
>>>>> Cheers,
>>>>>   Trond
>>>>>
>>>>> _______________________________________________
>>>>> nfsv4 mailing list
>>>>> nfsv4@ietf.org
>>>>> https://www.ietf.org/mailman/listinfo/nfsv4
>>>>>
>>>>>
>>>>>            
>>>>
>>>>
>>>> -- 
>>>> Best Regards
>>>> Sorin Faibish
>>>> Corporate Distinguished Engineer
>>>> Unified Storage Division
>>>>
>>>>        EMC²
>>>> where information lives
>>>>
>>>> Phone: 508-435-1000 x 48545
>>>> Cellphone: 617-510-0422
>>>> Email : sfaibish@emc.com
>>>> _______________________________________________
>>>> nfsv4 mailing list
>>>> nfsv4@ietf.org
>>>> https://www.ietf.org/mailman/listinfo/nfsv4
>>>>          
>>>
>>>        
>>
>>
>>      
> _______________________________________________
> nfsv4 mailing list
> nfsv4@ietf.org
> https://www.ietf.org/mailman/listinfo/nfsv4
>
>    