Understanding QFabric Multicast Flow Groups
In a QFabric system, a hash function is used to select an Interconnect device to forward traffic between two Node devices. Since this hash function is performed on all Interconnect devices, it is possible for redundant multicast streams to flow through one Interconnect device, making that Interconnect device a potential single point of failure for the redundant flows. Some applications, such as financial transactions, require that the redundant multicast streams flow through different Interconnect devices to prevent a single Interconnect device from potentially dropping both streams of multicast traffic during a failure. You can enforce this use of diverse Interconnect devices by using the QFabric flow groups feature.
For an explanation of QFabric system, see QFabric System Overview.
How Do I Implement a Flow Group in a Fabric?
There are two ways to configure a QFabric system, from the CLI and from Network Director. Flow control using flow groups for QFabric systems, however, can be configured from the CLI only, using the flow-groups statement.
You create a flow group from the CLI by designating the name of the flow group, the Node devices used by the flow group, and the Interconnect devices used by the nodes in the flow group.
When you configure a flow group, the software assigns a priority to each used Interconnect device. You can override the software settings using the CLI.
Do Flow Groups Impact QFabric System Performance?
Flow group configuration does not affect L3 multicast scaling numbers or performance.
What Platforms Support Flow Groups?
32 bit QFabric-G and QFabric-M platforms support flow groups. For more information about these platforms, see Understanding QFX3000-G QFabric System Hardware Configurations and Understanding QFX3000-M QFabric System Hardware Configurations.
How Do Multicast Flow Groups Work?
Link preferences, which restrict the traffic from ingress Node devices to Interconnect devices, are set to normal by default when they are not yet a part of a flow group. When you create a flow group, the link preferences between indicated Interconnect devices and Node devices are automatically changed from normal to either high or never, meaning that any preference determined by the normal algorithm is ignored and the flow is determined by the flow group link settings.
You can override the link preference by using the preference option of the flow-groups CLI statement.
The three link preferences used by the fabric software are:
high—Use the indicated Interconnect device in the hashing algorithm for the specified flow.
normal—Use the indicated Interconnect device for backup when Interconnect devices set to high fail.
never—Do not use the indicated Interconnect device in the hashing algorithm for the specified flow.
Interconnect device preference settings are assigned by the software. When you configure a flow group, preferences for some flows are set to high and some are set to never. This is how flows are enforced. Unassigned Interconnect devices (and Node devices) all belong to a default group and are set to normal. When an assigned Interconnect device fails, the corresponding flow looks for a normal Interconnect device to use as backup if one is available. This helps prevent loss of data in the event of an Interconnect device failure.
What are QFabric Multicast Flow Group Limitations?
An Interconnect device can belong to only one flow group.
A QFabric system can contain up to four flow groups.
By default, Node devices in a flow group cannot use Interconnect devices in other flow groups. You can override this setting using the preference option of the flow-groups CLI statement.