Network Load Balancing - jasper-zanjani/Windows-Server GitHub Wiki

Cluster VMs can be configured to drain their workloads to other nodes when being shutdown using Suspend-ClusterNode

NLB Clusters are made of hosts, while Failover Clusters are made of nodes.

NLB port rules control how the cluster functions and are defined by two operational parameters:

Affinity: associate client requests to cluster hosts. When no affinity is specified, all network requests are load-balanced across the cluster without regard to their source.
Filtering mode: specify how the cluster handles traffic described by port range and protocols; can be single or multiple hosts.

When a port rule is not configured, the default host will receive all network traffic.

Windows Server NLB Clusters can be upgraded to Windows Server 2016 in two ways:

Rolling upgrade brings only a single host down at a time, upgrading it before adding it and proceeding to the next one
Simultaneous upgrade brings the entire NLB cluster goes down

NLB clusters have a Cluster Operation Mode setting specifying what kind of TCP/IP traffic the cluster hosts should use

Unicast: NLB replaces the MAC address on the interface with the cluster's virtual MAC address, causing traffic to go to all hosts. Cluster hosts are prevented from communicating with each other in this mode. In this case, a second network adapter must be installed in order to facilitate normal communication between NLB cluster hosts.
Multicast: NLB adds a multicast MAC address to the network interface on each host that does not replace the original.