Steady-State Load versus Peak Load
When selecting a baseline configuration to use as a starting point for your ThingWorx deployment, it is important to understand what your steady-state load profile will be, but also plan for events that could cause spikes in load above that steady-state load level.
Some examples of peak load events could include a platform restart (all devices try to connect at once) adding or removing a node from a ThingWorx Cluster, a network outage recovery, an offline group of devices reconnecting and reporting past activity, adding a new group of devices, or adding new properties or measurements to an existing group of devices.
If you size your system to only handle steady-state load without accounting for potential spikes based on your use-cases, such a spike could lead to poor performance, data loss or unexpected outages.
In a high-availability configuration, it is also important to make sure that your steady-state load can continue to operate if part of the cluster were to go offline. In many cases, this will mean sizing your implementation so that steady-state load can be handled, but then adding one or more additional cluster nodes based on redundancy and peak load management requirements.