RE: iFCP as an IP Storage Work Item

[Dan_McConnell@Dell.com]

Regardless of native end to end IP SCSI transports, it is clear that we need
to address the existing "legacy" Fibre Channel devices (I still hesitate
calling Fibre Channel legacy, but that's a different story).  So, there is
an obvious need for bridging the gap between current FC devices and the
emerging IP storage networking realm.  On one hand we have a standard
tunneling protocol (FCIP) which would require the use of 1 FC switch port
and 1 FCIP "link extender" port on each side for each remote tunnel desired.
On the other hand the, iFCP equivalent would only require the use of 1 port
on the iFCP gateway (not per tunnel) which has the ability to route sessions
individually to the appropriate end iFCP gateway.  Assuming that the E-port
and L-port problems can be fixed, which supposedly should be no problem,
iFCP definitely seems to be a viable solution.  Whether iFCP or FCIP is the
functionality winner or not, I would say that it would be a loss for the IPS
working group to discard iFCP.  Also, the iSNS protocol, compatible with
both iSCSI and iFCP, has been accepted as a base for the iSCSI name service
and it seems to me that using iFCP would provide a perfect match for future
native iSCSI solutions to incorporate legacy FC devices over IP.  While FCIP
does address the issue of extending a FC link over IP,  I believe that iFCP
has added ip networking functionality and flexibility that is worth while
for the IPS working group to incorporate.

...that's my 20 centavos.....

Dan McConnell
Storage Systems Group
Storage Architecture and Technology Evaluation
Dan_McConnell@Dell.com