Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?

 
ON THIS PAGE
 

VCF Objective Metrics

This section provides information on the following objective metrics that you should consider when configuring a VCF:

Latency

VCF has a deterministic latency for both unicast and multicast traffic.

For unicast traffic forwarding, VCF utilizes the Djikstra shortest path first (SPF) algorithm. The overall latency is determined by the number of hops on the shortest paths from the ingress node to the egress node.

Multicast latency is determined by:

  • The multicast distribution tree (MDT) that is selected to forward a multicast packet. MDT selection is based on the multicast group’s next hop.

  • The distance—the number of hops—from the ingress node on the MDT to the egress node on the MDT.

Fabric Link Resiliency

When designing a VCF topology, you should ensure fabric link resiliency between any pair of nodes within the VCF to avoid cases of link and node failures. Fabric link resiliency is improved in a VCF by adding spine devices that are interconnected to all satellite devices in the VCF. A leaf device in a VCF that has four spine devices, for instance, has four paths to every other leaf device in the VCF.

You must configure at least two spine devices in all VCF topologies. A two spine device topology ensures some fabric resiliency as well as support for NSSU. In most VCF topologies, however, you should configure four spine devices if possible to provide maximum fabric link resiliency.

You do not want a VCF to split in the event of a failed link or member switch. A good design principle for fabric link resiliency is to ensure that, under any single link or switch failure, the VCF topology remains fully connected.

Fabric Bandwidth

When designing a VCF, know the total bandwidth of a particular VCF path and determine whether the topology design satisfies your traffic needs.

  • Unicast bandwidth—The maximum fabric bandwidth for unicast traffic between any pair of two nodes is the sum of the smallest segment bandwidth on all end-to-end paths. Therefore, unicast bandwidth between any node pair can be increased by adding intermediate devices in between. In a spine and leaf topology, bandwidth between leaf devices can be increased by adding spine devices.

  • Multicast bandwidth—Multicast traffic travels along a multicast distribution tree on the VCF topology. An MDT is composed of hop-by-hop link segments. Therefore, a VCF’s fabric multicast bandwidth is the minimum bandwidth of all aggregated fabric links between directly connected neighbors.