A speed mismatch occurs when the cluster ports on different nodes have different speeds, such as 10 Gbps and 40 Gbps1.
A speed mismatch can cause packet loss, retransmits, and latency on the cluster network, especially when data is accessed indirectly, meaning that the node serving the data is different from the node hosting the LIF2.
In this scenario, the AFF A300 nodes have 10 Gbps cluster ports, while the AFF A700 nodes have 40 Gbps cluster ports, creating a speed mismatch on the cluster interconnect34.
The Cisco Nexus 3132Q cluster interconnect switches support both 10 Gbps and 40 Gbps cluster ports, but they have a shallow buffer size of 9 MB per port, which can be easily overwhelmed by the bursty traffic generated by the speed mismatch.
To solve this problem, you need to either upgrade the cluster ports on the AFF A300 nodes to 40 Gbps, or downgrade the cluster ports on the AFF A700 nodes to 10 Gbps, to achieve a homogeneous cluster interconnect1.
Alternatively, you can use QoS policies to limit the bandwidth of the cluster ports to avoid oversubscription and congestion on the cluster interconnect.Reference:
1: Cluster network cabling, ONTAP 9 Documentation Center
2: Cluster Network Latency - Troubleshooting Guide, NetApp Knowledge Base
3: AFF A300 Tech Specs, NetApp
4: AFF A700 Tech Specs, NetApp
[5]: Why do network speed mismatches create problems with shallow buffered switches?, NetApp Knowledge Base
[6]: How to use QoS to limit the bandwidth of cluster ports, NetApp Knowledge Base
