Re: FCIP: RE: iFCP

["Franco Travostino" <travos@nortelnetworks.com>]

At 07:15 AM 11/28/00, Wayland Jeong wrote:

[ stuff deleted ]

When islands of SAN are connected via FCIP devices, two routing planes are
created: the IP routing plane and the FC routing plane. The FC routing
protocol, FSPF, knows nothing about how a FC frame encapsulated in IP gets
from one FCIP device to another. To FSPF, it simply looks like an ISL
between E-ports. Thus, when an FCIP device advertises its link state on the
FC network (per SW), it will publish its next hop connections in terms of
domains on the other side of an IP connected FCIP device. Frames are routed
in the IP network using standard IP routing semantics like OSPF. If an FCIP
device has multiple IP egress spigots, then it will be able to make the same
next hop decisions like any other IP forwarding device.

Thanks, Wayland. The FSPF routing event would result in a different egress spigot and a different FCIP tunnel, and the latter needs to be manually provisioned (ahead-of-time or just-in-time). Whereas iFCP's doctored {D_ID, S_ID} tuple in the FC frame would continue to hold and identify the state at the recipient gateway regardless of the egress spigot.

As for composing FC routing plane and IP routing plane ... I notice that in your scenario the FSPF Hello is a virtual hop-by-hop signalling and no longer a physical hop-by-hop signalling (i.e., it may now be tunnelled through gobs of IP hops along metro or even wan distances). Unduly tight FSPF Hello intervals (tunable, seconds resolution) will make FSPF flap even over benign IP transient failures. This is not a problem that a 10 minutes Hello interval cannot fix. Some good multiples of RTT are probably in order. One routing plane is plenty enough if I can help it :-)

-franco

-----
Franco Travostino, Director Content Internetworking Lab
Technology Center
Nortel Networks, Inc.
600 Technology Park
Billerica, MA 01821 USA
Tel: 978 288 7708 Fax: 978 288 4690
email: travos@nortelnetworks.com