Compatible mapping between servers for failure recovery.

["Douglas Otis" <dotis@sanlight.net>]

All,

To expand on the concept of creating a universal map to ensure compatibility
between servers in the access of underlying SCSI structures, here is an
example of a possible mapping.  In the event of an iSCSI server failure, a
different server would allow identical access using the same map.  The
exception would be the Service Delivery Port that could be unique to the
particular server.

 0                   1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|   MSB LUN     |   LSB LUN     |MSB Service Delivery Port (SDP)|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|  LSB (SDP)    |                 Target Port                   |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
                     iSCSI LUN Mapping

In the case of Fibre-Channel, the Target Port would be the D_ID for the FC
port.  Should the target represent a parallel configuration, the two byte
LUN would contain the address specific method for logical unit addressing.

  0                                       1
  0   1   2   3   4   5   6   7   8   9   0   1   2   3   4   5
+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
| 1   0 |        Target         |    Bus    |       LUN         |
+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
            MSB/LSB LUN breakdown for parallel SCSI

If there is a write cache, these caches must be synchronized between servers
prior to sending a completion response to the client.  Only one server would
have immediate access to the owned set of the underlying storage.  If both
servers are operational, each would be operating from an exclusive set of
devices.

Creating this type of convention allows fewer unique states to be present
within the iSCSI server and thus reducing the number of unknowns resulting
from a server failure.  Use of external authentication servers deliver to
the client the permitted LUNs via a list.  This list would simply be an
array of N number of iSCSI LUN structures.  Should a server fail, access via
a redundant server should be possible by revisiting the authentication
server and obtaining the information for the currently operational iSCSI
server.

In the future, it may be practical to use the two byte LUN field to replace
partitions.  Should only portions of a drive be granted access, a partition
affords little protection and with the security features of the SCSI portal,
the separate partitions would be maintained as an isolated LUN restoring a
possible use for LUNs.  As this would take time to instantiate should there
be a demand, such use should only be a consideration.

Doug