[Ksummit-2008-discuss] Kernel Summit Request for Discussion: The Future of Target mode and Cloud storage on Linux

Nicholas A. Bellinger nab at linux-iscsi.org
Wed Aug 20 19:21:04 PDT 2008


On Thu, 2008-08-21 at 11:04 +0900, Paul Mundt wrote:
> On Wed, Aug 20, 2008 at 06:43:34PM -0700, Nicholas A. Bellinger wrote:
> > On Thu, 2008-08-21 at 10:40 +0900, FUJITA Tomonori wrote:
> > > On Wed, 20 Aug 2008 18:26:13 -0700
> > > "Nicholas A. Bellinger" <nab at linux-iscsi.org> wrote:
> > > 
> > > > > > Yes there is another, IET, the one that has been included into a few
> > > > > > distributions, and who's iSCSI logic has been included into other
> > > > > > projects.
> > > > > 
> > > > > I think that Debian and Suse have already replaced IET with STGT.
> > > > > 
> > > > > 
> > > > 
> > > > Well, are redhat.com and suse.com services running on top of fabric from
> > > > either IET or STGT yet..?  Linux-iSCSI.org services are, as will be
> > > > running on top of open source storage cloud as we move towards VHACS
> > > > v1.0.
> > > 
> > > Sorry, I don't understand what you mean. Why does redhat.com and
> > > suse.com services need to run on 'cloud' technology?
> > > 
> > > You said that IET has been included into some distributions so I just
> > > pointed out that some of them include STGT instead of IET now.
> > > 
> > 
> > Sorry Tomo-san, it was a small jab at the fact that STGT is not being
> > used recommended for production usage, and I am guessing that those
> > websites would rather run on fancy storage arrays with closed IP storage
> > stacks (I had no idea what they are using btw), instead of running on
> > top of their own target mode stack that they ship (as Linux-iSCSI.org
> > does).
> > 

Hi Paul,

> Yes, one could almost start to think that these are distribution vendors
> that have their feature requirements determined by customers, or
> something equally obscure. I'd be more concerned if random distro vendors
> started implementing their own target mode stacks just for the hell of
> it.
> 
> If you don't know what they're running, then what was the point of even
> bringing it up? It's been pointed out that there are existing STGT users
> that are shipping today and now you are trying to suggest that what they
> run their .com on is in some way relevant to the discussion at hand,
> whether or not they happily ship and support it already.
> 

The point is that they are not recommending it for production, but
including it for people to use.  End of story.

> This is really the wrong forum for empty rhetoric, and it's also not
> going to help you make your case, especially to an audience of people who
> have no idea what you're going on about.
> 
> The fact STGT exists today and it seems that there's interest in
> developing the kernel-side of things if someone puts the work in to it is
> a pretty compelling argument for not ripping it out completely and simply
> finding ways to improve what's already there. Unfortunately this tends to
> be fairly incompatible with the my-stack-is-better-than-yours approach to
> kernel development favoured by so many, especially when those people take
> their ball in to their own corner and show up some time later with a 50k
> code drop.
> 

That is not what I am advocating, what I am advocating are the steps to
get to where real production usage of target mode on Linux using
upstream code can become a reality.

--nab




More information about the Ksummit-2008-discuss mailing list