Layer 3 multicast load balancing establishes a more efficient distribution of Layer 3 multicast data over ECMP and LAG links. Operators have the option to redistribute multicast groups over ECMP or LAG links if the number of links changes either up or down.
When implementing this feature, there are several considerations. When multicast load balancing is not configured, the distribution remains as is. Multicast load balancing is based on the number of ‟s,g” groups. This means that bandwidth considerations are not considered. The multicast groups are distributed over the available links as joins are processed. When link failure occurs, the load is distributed on the failed channel to the remaining channels so multicast groups are evenly distributed over the remaining links. When a link is added (or failed link returned) all multicast joins on the added links are allocated until a balance is achieved.
When multicast load balancing is configured, but the channels are not found in the multicast-info-policy, then multicast load balancing is based on the number of ‟s,g” groups. This means that bandwidth considerations are not considered. The multicast groups are distributed over the available links as joins are processed. The multicast groups are evenly distributed over the remaining links. When link failure occurs, the load is distributed on the failed channel to the remaining channels. When a link is added (or failed link returned) all multicast joins on the added links are allocated until a balance is achieved. A manual redistribute command enables the operator to re-evaluate the current balance and, if required, move channels to different links to achieve a balance. A timed redistribute parameter allows the system to automatically, at regular intervals, redistribute multicast groups over available links. If no links have been added or removed from the ECMP/LAG interface, then no redistribution is attempted.
When multicast load balancing is configured, multicast groups are distributed over the available links as joins are processed based on bandwidth configured for the specified group address. If the bandwidth is not configured for the multicast stream then the configured default value is used.
If link failure occurs, the load is distributed on the failed channel to the remaining channels. The bandwidth required over each individual link is evenly distributed over the remaining links.
When an additional link is available for a specific multicast stream, then it is considered in all multicast stream additions applied to the interface. This multicast stream is included in the next scheduled automatic rebalance run. A rebalance run re-evaluates the current balance with regard to the bandwidth utilization and if required, move multicast streams to different links to achieve a balance.
A rebalance, either timed or executing the mc-ecmp-rebalance command, should be administered gradually to minimize the effect of the rebalancing process on the different multicast streams. If multicast re-balancing is disabled and subsequently enabled, keeping with the rebalance process, the gradual and least invasive method is used to minimize the effect of the changes to the customer.
By default multicast load balancing over ECMP links is enabled and set at 30 minutes.
The rebalance process can be executed as a low priority background task while control of the console is returned to the operator. When multicast load rebalancing is not enabled, then ECMP changes are not be optimized, however, when a link is added occurs an attempt is made to balance the number of multicast streams on the available ECMP links. This however may not result in balanced utilization of ECMP links.
Only a single mc-ecmp-rebalance command can be executed any specific time, if a rebalance is in progress and the command is entered, it is rejected with the message saying that a rebalance is already in progress. A low priority event is generated when an actual change for a specific multicast stream occurs as a result of the rebalance process.