Re: [nfsv4] List of possible work items for NFSv4.2

[sfaibish <[e-mail]>]

Mike,

I was in vacation the last 2 weeks and I couldn't reply to your email until
now. I hope it is not too late.

On Sat, 15 Aug 2009 10:58:06 -0400, Mike Eisler <[e-mail]>  
wrote:

> Reading through responses, both those published to the list and privately
> sent to me, I see a strong consensus for NFSv4.2, strong indication
> that there are people willing to contribute to the document process as
> orginators, editors, and reviewers, and strong indication that this work
> will be prototyped/implemented.
>
> I propose:
> - that NFSv4.2, RPCSEC_GSSv3, MAC be authorized as a work items  
> immediately.
> -- I further propose that the primary editors of these three work items  
> be,
>    respectively, Mike Eisler, Nico Williams, and Dave Quigley.
I will help with the review.

> - For the NFSv4.2 document, we start by combining the COPY operations  
> I-D,
>   and the part of the MAC I-D that describes the RECOMMENDED MAC  
> attribute.
> -- We should have a discussion and reach consensus on how to combine
>    sub-file caching
> (http://tools.ietf.org/html/draft-eisler-nfsv4-pnfs-dedupe
>    or
> http://tools.ietf.org/html/draft-myklebust-nfsv4-byte-range-delegations),
>    peer-to-peer NFS (
> http://tools.ietf.org/html/draft-myklebust-nfsv4-pnfs-backend ),
>    and unstable file create (see Trond's emails on this thread),
>    and then fold this consensus into NFSv4.2.
I will be part of the discussion. I would also be interested to combine
the sub-file caching for file, block and object layouts with some cross
mapping. I can explain in more detail if it is of interest. I would also
add to this perhaps an option to add later additional layout drivers. I  
will
be also interested to extend the P2P to take care of large numbers of small
files to address the explosion of number of objects scalability. I believe
that Benny Halevy will be also interested to contribute.

I want also to propose including a tighter error detection and reporting
related to pNFS access permissions I proposed at the meeting. I will extend
similar stats permission errors for the pNFS server.

One additional item should be the support of IPV6 in pNFS if there are any
gaps in the protocol. I will try to work with Chuck Lever on this item.

>
> Going through the items one by one:
>
> On Wed, August 12, 2009 8:41 pm, Mike Eisler wrote:
>
>> Peer-to-Peer NFS
>
> Strong indication of support for this work in terms of reviewing,
> contributing and implementing (it has already been prototyped).
> Trond wants to re-do the prototype once the NFSv4.2 form of this
> work is clarified.
[Sorin] I will help with the review and test the prototype.

>
>> Copy
>
> James indicated he'd like to continue to contribute here, including
> with implementation. Tom Haynes also indicated a possibility to  
> contribute
> here.
>
>> Hole Punching
>
> I received a private response that proposals for hole punching and
> thin provisioning will be coming soon.
[Sorin] I will contribute to this item including possible prototyping.

>
>> MAC Security Attribute
>
> Dave Quigley is committed to contributing to
>
>> Traffic Classification
>
> Thus far, no renewed interested has been expressed on this.
>
>> End-to-End Data Integrity
>
> Thus far, no renewed interested has been expressed on this.
>
>
>> Umask Attribute
>
> Thus far, no renewed interested has been expressed on this.
>
>> Shutdown Callback
>
> Thus far, no renewed interested has been expressed on this.
>
>> Readahead Hint
>
> Thus far, no renewed interested has been expressed on this.
>
>> pNFS Connectivity/Access Indication
>
> A fresh item from Stockholm IETF meeting; absence of
> re-stating support for this is not absence of proof of interest.
> I for one am committed to seeing this solved in NFSv4.2.
[Sorin] I will be involved in drafting the changes.

>
>> Better Negotiation of Session Reply Cache Sizes
>
> A fresh item from after Stockholm IETF meeting; absence of
> re-stating support for this is not absence of proof of interest.
> But I remain luke warm on this ... not entirely convinced there is
> a problem to solve.
>
>> Metadata Striping
>
> Andy Adamson has expressed interest in prototyping this.
[Sorin] This relates close to the large number of small objects scalability
and I will work with Andy on this.

>
>> De-Dupe Awareness and Sub-File Caching
>
> In the context of peer-to-peer NFS, I see continued interested
> in sub-file caching. Sub-file caching as proposed my me, gives
> one de-dupe awareness for free (or the converse depending on
> how one wants to look at it). As I've proposed at the beginning of
> this message, the WG needs to discuss how it wants to approach all
> these topics before adding them to an NFSv4.2 I-D.
[Sorin] I want to contribute to this.

>
> New items:
>
> - unstable file creates. This has a lot of interested within the WG
> - capabilities: An item proposed by Nico Williams
> - attributes: Dan Muntz expressed an interest in revisiting how  
> attributes
>   are handled.
>
>

/Sorin


-- 
Best Regards

Sorin Faibish
Corporate Distinguished Engineer
Network Storage Group

         EMC²
where information lives

Phone: [phone] x 48545
Cellphone: 617-510-0422
Email : [e-mail]