Portchannels; Virtual Port Channel Host Mode; Load Balancing; Source-Based Hashing - Cisco Nexus 1000V Series Deployment Manual

Deployment Guide
In some scenarios, single standard uplinks may be used, such as for a VMware ESX host that contains only a single
NIC. Another possible scenario is a host connected to a secondary network that does not require high availability,
such as a single NIC connected to a dedicated backup network. Given the requirements of most data center
networks, however, standard uplinks should rarely, if ever, be used in a Cisco Nexus 1000V Series design.

PortChannels

The Cisco Nexus 1000V Series implements a PortChannel mechanism that supports two modes; standard
PortChannels and virtual PortChannel Host Mode (vPC-HM). Regardless of the mode, PortChannels are managed
using the standard PortChannel CLI construct, but each mode behaves differently.
A standard PortChannel on the Cisco Nexus 1000V Series behaves like an EtherChannel on other Cisco switches
and supports the Link Aggregation Control Protocol (LACP). Standard PortChannels require that all uplinks in the
PortChannel be in the same EtherChannel on the upstream switch.
Standard PortChannels can be spread across more than one physical switch if the physical switches are clustered.
®
Examples of clustered switching technology include the Catalyst 6500 Virtual Switching System 1440, virtual
PortChannels on the Cisco Nexus 7000 Series Switches, and the Cisco Catalyst Blade Switch 3120 for HP. Clustered
switches act as a single switch and therefore allow the creation of EtherChannels across them. This clustering is
transparent to the Cisco Nexus 1000V Series.

Virtual Port Channel Host Mode

Most access layer switches do not support clustering technology, yet most Cisco Nexus 1000V Series designs
require PortChannels to span multiple switches. To enable this spanning of switches, you can use a PortChannel in
vPC-HM. vPC-HM divides the PortChannel into subgroups, with each subgroup representing one or more uplinks to
one upstream physical switch.
Note:
When vPC-HM is used, the Cisco Nexus 1000V Series supports only two subgroups: 0 and 1.
Links within the PortChannel that are connected to the same physical switch are bundled in the same subgroup
automatically by using the Cisco Discovery Protocol packets received from the upstream switch. Alternatively,
interfaces can be manually assigned a specific subgroup using interface-level configuration.
When vPC-HM is used, each vEth interface on the VEM is mapped to one of the two subgroups using a round-robin
mechanism. All traffic from the vEth interface uses the assigned subgroup unless the assigned subgroup is
unavailable, in which case the vEth interface will fail over to the remaining subgroup. When the originally assigned
subgroup becomes available again, traffic will shift back to its original location. Traffic from each vEth interface is then
hashed within its assigned subgroup based on the configured hashing algorithm.
vPC-HM should be considered the preferred method for most Cisco Nexus 1000V Series deployments.

Load Balancing

The Cisco Nexus 1000V Series provides 17 hashing algorithms to load-balance traffic across physical interfaces in a
PortChannel. These algorithms can be divided into two distinct categories: source-based hashing and flow-based
hashing.
Select the hashing algorithm used carefully as it affects the available configuration options and may require
configuration changes on the access layer switches. The default hashing algorithm used by the Cisco Nexus 1000V
Series is source MAC address hashing (a source-based hash).

Source-Based Hashing

Source-based hashing algorithms help ensure that a MAC address is transmitted down only a single link in the
PortChannel, regardless of the number of links in a PortChannel. When you use source-based hashing, upstream
switches are still required to use EtherChannel when connecting to a VEM.
© 2009 Cisco Systems, Inc. All rights reserved. This document is Cisco Public Information. Page 18 of 25