P1394 Mail Archive: Re: P1394> Dynamic LUNs

P1394 Mail Archive: Re: P1394> Dynamic LUNs

Re: P1394> Dynamic LUNs

Greg Shue (gregs@sdd.hp.com)
Tue, 26 May 1998 09:36:44 -0700 (PDT)

SBP-2 says that LUN zero must exist.

When PWG representatives went to Redmond, Microsoft said that
their driver only supports LUN zero. They also said that LUN
zero needs to be directly mapped to a higher layer, because they
won't support dynamic LUNs.

So, we should think of LUN zero as being "allocated".

NOTE: Someone (Brian?) in VA. said that Microsoft's driver
will be extended to support more LUNs. There was a comment made
(from the Microsoft rep?) that Microsoft still will not support
dynamic LUNs.

> I'm not sure what you are recommending in comment (B) below. If this came
> up at the Virginia meeting, can someone elaborate as if LUN 0 for a
> particular unit has been "allocated" by Microsoft or some other group?
> Randy
> At 05:03 PM 5/14/98 -0700, Greg Shue wrote:
> >
> >Here's some comments on the Dynamic LUN notes sent out by Alan.
> >
> >A) I don't think the Dynamic Logical Unit Number scheme was
> > intended to _multiplex_ applications to device functions.
> > I think it is intended on providing the ability for a node
> > to establish a unique connection to identical instances of
> > a service without having to:
> >
> > 1- Assign fixed LUNs for each service instance
> > 2- Crowd the Config ROM with redundant information
> >
> >B) After discussions with folks from Microsoft, I would strongly
> > recommend NOT using LUN 0 (zero) for anything but a direct
> > connection to an instance of the transport client. The LUN-server
> > could easily be identified in the Config ROM as belonging to
> > a different LUN. (Since it's really a different service
> > than the transport client.)
> >
> >C) The SBP-2 initiator and target drivers do not change a bit.
> > Reconnect proceeds normally. The initiator and target SBP-2
> > drivers must already remember LUNs, EUI-64 values, and LoginIDs.
> >
> >D) The SBP-2 driver probably will already map connections (LoginIDs?)
> > to a socket ID or some other appropriate OS mechanism.
> >
> >It still strikes me as a clean, simple way to get beyond the constraints
> >of SBP-2 when appropriate. Best of all, it's optional!
> >
> >--
> >Greg Shue
> >Hewlett-Packard Company
> >Office Products Division gregs@sdd.hp.com
> >----------------------------------------------------------------
> >

Greg Shue
Hewlett-Packard Company
Office Products Division			gregs@sdd.hp.com