Like Juniper, Cisco relies on global repair and local repair to assure Link
Protection. The global repair is solely based on a RSVP PathErr message
going back to the ingress router (with error code of "Notify" and an error
value of "Notification of local repair"). Once received, the ingress takes
appropriate actions to create a new LSP.
For local protection, Cisco uses a different mechanism then Juniper. This
mechanism is described in:
draft-swallow-rsvp-bypass-label-01
What is called detour in Juniper terminology is here called a bypass tunnel.
The main difference between bypass tunnel and detour is that the bypass
tunnels are not associated with any LSP. Bypass tunnel are associated and
created to bypass a link that may fail. Every LSP that is using a particular
link between two routers may use the bypass tunnel created for that given
link. This, as the draft mention, improves the scalability of the mechanism
since every LSP does not have to establish its own bypass tunnels. Since
many LSP may use the same bypass tunnel, it is important to have a means for
the merge node to know to which LSP a particular packet is belonging to. To
solve this, the label associated with the bypass tunnels is stacked on top
of the original label. The stacked label is removed by the penultimate hop
of the bypass tunnel so that the merge node only sees the label it was
expecting to receive before the failure occurred.
>
If the node of local repair, R2, and merge node, R3, were not directly
connected to each other (another router between the two), R2 would not know
which label it should use to communicate with R3 since labels are only
significant for a given link. To solve that, the draft uses the label record
option of the ROUTE_RECORD object. With that option, upstream nodes could
learn about downstream labels not directly connected to the same link.
Label space is also a concern. Let suppose that R3 has a per interface label
space. At the creation or the primary LSP, R3 binds a label based on the
interface the LSP is coming from. But, if the one link fails and the bypass
tunnel becomes active the same label will then arrived by another interface.
Since R3 has a per interface label space, it will not be able to recognize
it as a valid label or worst, the packet will be sent through the wrong LSP.
The draft is unclear about how this should be solved. This is my first
question the second is according to cisco the bypass tunnel is standby
statis the problem I can't understand is RSVP uses the local IGP to setup
the LSP. OSPF does not have the capabilities to see 2 paths it can only see
its next hop. Did they modify ospf to recompute a route while blocking R3
from being the next hop? Or is the 2nd tunnel setup another way?
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=4559&t=4559
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
