Flow-Based Hashing; Network-State Tracking - Cisco Nexus 1000V Deployment Manual

Switch for microsoft hyper-v

Hide thumbs Also See for Nexus 1000V:

Table of Contents

The following Cisco Nexus 1000V algorithms can be classified as source-based hashes:

Virtual port ID

●

Source MAC address

●

Flow-Based Hashing

Flow-based hashing enables traffic from a single MAC address to be distributed down multiple links in a

PortChannel simultaneously. Use of a flow-based hash increases the bandwidth available to a virtual machine or

to Live Migration and increases the utilization of the uplinks in a PortChannel by providing more precise load

balancing.

Flow-based hashing algorithms are any algorithms that use the following to hash:

Packet destination

●

Layer 4 port

●

Combinations of source address, destination address, and Layer 4 port

●

Network-state tracking (NST) is a mechanism that is used to detect Layer 1 and network connectivity failures that

would otherwise cause virtual machine traffic to be dropped during an uplink failure. An uplink failure occurs when

the upstream switch encounters a driver or firmware failure that prevents it from signaling a link to the neighboring

link interface (the Cisco Nexus 1000V VEM in this case). NST mitigates this problem by using NST packets, which

probe interfaces on other subgroups of the same VEM.

When a link failure occurs on the upstream switch, NST can detect the failure by sending the tracking packet from

one interface in Subgroup 0 to all interfaces in Subgroup 1. Because a link failure occurred on the upstream

switch, this tracking packet will not be received by Subgroup 1. The vPC-HM PortChannel interface will then be

identified as split, and a syslog will be generated for this split. Also, packet counters are monitored to detect

whether there is traffic coming into the subgroup that did not receive the tracking packet. If packet counters are not

incrementing, then the upstream switch is not transmitting to the VEM in the subgroup. At this point, traffic that

was originally pinned to the Subgroup 1 interface will be repinned to the active interface in Subgroup 0.

You should enable NST when an uplink failure may cause traffic to be dropped. The most common deployment of

NST is in cases in which third-party networking entities may encounter driver or firmware or other software failures

in which the interface is not brought down completely and the VEM interface continues to forward traffic.

Page 46 of 48

Table of Contents