Showing results for 
Search instead for 
Did you mean: 

Shortest path with ISLs?

Shortest path with ISLs?

New Contributor

Sorry if it has been cleared before, but is there any short-path intelligence with ISL links?

We have a fabric of six 6740T-1G VDXs running NOS 6.0.2b in a triangle (L2, stretched), and would like to ensure that when the server from Switch1 talks to the server from Switch2 it doesn't traverse via Switch3.

Thank you.

Extreme Employee
Correct hop count is effectively the only metric used

Extreme Employee
The protocol FSPF (Fabric Shortest Path First) will be used and this will ensure that the path with the least cost will be used.

Of all of links should have the same cost therefore the single hop link will always be preferred as it will have the lowest cost.

Extreme Employee
Unfortunately there is no any short-path intelligence with ISL links

VCS uses the Fabric Shortest Path First (FSPF) routing protocol to distribute link-state information of all ISLs. FSPF is a Link State Path Selection protocol, similar to OSPF, which is an Interior Gateway Protocol (IGP) widely used in IP networks.

FSPF protocol keeps track of the state of the links on all switches in the Ethernet fabric. It also associates a cost with each link. The protocol computes paths from a switch to all the other switches in the fabric, by adding the cost of all the links traversed by the path, and choosing the path that minimizes the cost.

FSPF is similar to Layer 3 routing protocols like OSPF. Although it has roots from OSPF, FSPF only defines and implements point-to-point links. In other words, there is no concept of a designated router (DR) and a backup designated router (BDR), areas or summarization, or anything similar like that being managed in FSPF.

FSPF forms a single adjacency per fabric trunk.

FSPF Link Cost
ECMP in an Ethernet fabric behaves slightly differently from the traditional L3 ECMP. Link cost is a metric value assigned to the transmit (Tx) side of each ISL port. The link cost (metric) value for all interfaces is 500, regardless of bandwidth. A 10 Gbps interface has the same link cost as an 80 Gbps fabric trunk. If a neighbor switch is reachable through several interfaces of various link speeds, all of them are treated as ECMP routes.


as I understand your post the shortest path is used, but due to a fixed link cost the metric is practically just the hop-count between the end points.