Cluster Network Latency - Troubleshooting Guide

The term "cluster interconnect" or "cluster network" refer to the Ethernet network allowing nodes in a cluster to communicate with each other
- This network is used to provide a single client and administrative view of a multi-node cluster without
- When a client connects to a network interface on the cluster, it can access data in the Storage Virtual Machine (SVM) it has access to, irrespective of which node the data resides on
- File access for any request is forwarded to the correct node and returned over the cluster network.
For faster performance and resiliency, NetApp has expanded the number of size and speed of the cluster interconnect
- Currently on FAS/AFF platforms, from two to eight Ethernet ports are supported with speeds of 10GbE, 40Gb, and 100GbE

Homogenous cluster interconnect	Each node in the cluster has equal number of cluster Ethernet ports Each node's cluster ports are at equal speeds Example: A 4 node cluster has 4 10Gb Ethernet ports per node
Heterogenous cluster interconnect	A disparity exists between the number or speed of cluster ports If using a 40Gb Ethernet breakout cable, the speed is considered 10Gb not 40Gb Example: Some nodes may have 4 10Gb Ethernet ports, while some may have 2 40Gb Ethernet ports in a Cluster

Packet loss is seen and this causes huge delays on the cluster network
Please see the below articles for more explanation of why packet loss causes problems:
- Why do network speed mismatches create problems with shallow buffered switches?
- Why does packet loss impact performance?

Note: For more articles on other causes, see the article Cluster Interconnect Latency - Resolution Guide.