When you configure a cluster peer relationship, you assign intercluster LIFs to ports capable of handling intercluster communications. Before selecting the ports, you have to determine whether to use dedicated ports for intercluster communication or to share ports that are also used by the data network.
Network ports in a cluster have roles that define their purpose and their default behavior. Port roles limit the types of LIFs that can be assigned to a port. Intercluster LIFs use only data ports or intercluster ports.
You can assign any Ethernet port type in the system the role of data or intercluster, including physical ports (such as e0e or e0f) and logical ports (such as a VLAN or an interface group).
You can create VLANs and interface groups on data ports. VLANs and interface groups have the data port role by default.
When you assign the intercluster role to a port, Data ONTAP automatically configures a non-modifiable failover group consisting of intercluster ports for the intercluster LIFs. Data protocols cannot fail over or migrate to an intercluster port.
If you used Data ONTAP operating in 7-Mode and used a dedicated port for replication traffic, it is likely that you will want to use a dedicated network for clustered Data ONTAP. Most of the networking considerations you encountered when using replication in 7-Mode also apply to clustered Data ONTAP.
If you have a high-speed network, such as 10 GbE, you might have enough local LAN bandwidth to perform replication using the same 10 GbE ports used for data access.
If you have a 10 GbE network, you should compare your available WAN bandwidth to your LAN bandwidth. If the available WAN bandwidth is significantly less than 10 GbE, you might be limited to the network utilization that the WAN can support.
If you plan to use the peer relationship for replication, consider how your available bandwidth will handle the level of client activity during the replication interval. If the WAN bandwidth is similar to that of the LAN ports and replication will occur during regular client activity, then dedicate Ethernet ports for intercluster communication to avoid contention between replication and the data protocols.
Consider the amount of data that will be replicated in each interval and whether it requires so much bandwidth that it could cause contention with data protocols for shared data ports. If you use the peer relationship for replication and replication is set to occur only when minimal to no client activity occurs, you might be able to use data ports for intercluster replication successfully, even without a 10 GbE LAN connection.
If the network utilization generated by the data protocols (CIFS, NFS, iSCSI) is above 50%, then you should dedicate ports for intercluster communication to allow for non-degraded performance if a node failover occurs.
Sharing ports for data and intercluster communication eliminates the extra port counts required when using dedicated ports. When data ports for intercluster communication are shared, the intercluster LIFs can be migrated to any other intercluster-capable port on the same node.
When physical 10 GbE ports are used for both data and intercluster communication, you can create VLAN ports and use dedicated logical ports for intercluster communication.
Using dedicated ports for intercluster communication requires additional switch ports and cable runs.
For best performance, all paths used by intercluster LIFs should have equal performance characteristics. If a node has one intercluster LIF on a slow path and another intercluster LIF on a fast path, performance will be adversely affected because data is multiplexed across the slow and fast paths simultaneously.