Proposal for IEEE company_id based formats for FC-PH world-wide identifiers
mw at core.rose.hp.com
Fri Dec 20 14:20:58 PST 1996
* From the SCSI Reflector (scsi at symbios.com), posted by:
* Mike Wenzel <mw at core.rose.hp.com>
Thanks for the more complete version. After looking it over, I can't decide
how vital an "association" field would be (to give the RAID controller-to-
virtual LUN correspondence).
I've spent quite a bit of time recently looking at the entities with WWNs
for FibreChannel SCSI. Many of these names will be crucially important to
efficient operation, especially for FC Loop, where addresses could change
abruptly with any loop reinitialization--the port name and/or node name
provides the stability for resynchronization among nodes that the addresses
used to. There are some "head-scratcher" issues, though, about how long
the host should hang onto a port or node name, especially in the face of
on-line replacement of FRUs. Also, the topologies can quickly become very
complex--to the point where it's hard to say how significant it is for the
host computer to know what all the intermediate nodes are between itself
and a chunk of storage. Most names seem important for transient recovery
|from renegotiated addresses, and for finding and identifying failed FRUs.
For these purposes, it isn't obvious how far associations need to extend
beyond the named object itself.
However, one thing keeps coming back as being crucially important and an
anchor in all this business: the world-wide name for the chunk of storage
itself. Host computers can adapt to almost any "how" for reaching a
chunk of storage, as long as they know when they HAVE reached the same
chunk of storage! So the one invariant that I keep coming back to like a
broken record is for the WWN of the LUN (storage chunk) to be the same,
regardless of the paths used to reach it. I'm really glad that Bob
Sniveley's write-up covers this adequately.
In short, I'm not sure what discussions need to take place or what we need
to look at in order to decide how hard to pursue an association field.
Best Regards and Searons Greetings,
************* | Mike Wenzel,
***** ***** | Hewlett-Packard - NCD System Interconnect Lab,
*** /_ _ *** | Mailstop 5601,
** / / /_/ ** | 8000 Foothills Blvd.,
*** / *** | Roseville, CA 95747-5601
***** ***** | E-mail: mw at core.rose.hp.com
************* | Telephone (916) 785-5609 FAX (916) 785-2875
At 12:59 PM 12/19/96 PST, Larry Chen wrote:
>Assume the following configuration:
>Host Computer <----FC-AL----> RAID controller + disk drives + enclosure
>The host computer is attached directly to the RAID controller only
>(not the disk drives). The RAID controller will manage the disk drives
>and export virtual LUNs to the host computer.
>At the SCSI initiator's side and level, I wanted a mechanism to associate the
>virtual LUNs with its RAID controller. This mechanism would be general enough
>to work for both the simple disk drive case and for the RAID case.
>Below, I will attempt to describe the entities at the FC and SCSI layers
>for my current configuration.
>At the FC transport layer
>Port entity is a SIM in CAM terminology.
>Node entity is the RAID controller (or more specifically,
> the RAID controller's Field Replaceable Unit (FRU)).
>At the SCSI layer
>LUN entity is a virtual LUN.
>RAID controller entity is the FRU.
>In my single RAID controller configuration, the FC Node entity and the
>SCSI RAID controller entity are equivalent (and probably will have the same
>Port name - any old WWN
>Node name - IEEE Registered Format
>Virtual LUN name - IEEE Registered Extended Format
>RAID controller name - IEEE Registered Format
>and the association field could be used to associate the Virtual LUNs
>with its RAID controller.
>A problem arises when multi-RAID controller complexes are introduced.
>Now, the RAID controller entity is ambiguous (FRU vs. the virtual
>RAID controller) and the FC Node entity and the SCSI RAID controller
>entity are probably not the same.
>At this point, I will not be pursuing this matter any further (unless
>someone is willing to stand-up with me).
>As an observation, RAID devices seem much more complex when compared to
>the simple disk drive case. I hope the RAID Industry will form some
>initiatives that will rectify this soon.
>Please call me if there are any further questions.
* For SCSI Reflector information, send a message with
* 'info scsi' (no quotes) in the message body to majordomo at symbios.com
More information about the T10