Why am I getting CPU Overload errors in xPC Target even though my sample time is much larger than the TET's?

I have a model which has networking capabilities in it. Whenever I start it up it will error out on the second sample with a CPU Overload error. I also lose the TCP/IP communication between the target and the host. Why is this?

Answer 1

[MathWorks Support Team]

If a model heavily relies on the use of the network interface card (NIC) in the target, latencies introduced by this device play an important role. CPU overload errors can be caused by having too much (non-xPC) traffic sent to the target's NIC, which will make it's latency bigger, directly influencing the execution time.

To overcome this, make sure that this network device is only receiving data that is actually meant for the target.

There are several ways of achieving this:

1) Create a dedicated network or subnet for your xPC environment

2) Put the target machine behind a switch, so that the switch will take care of filtering out the traffic that is not meant to be sent to the target. Do not use a hub, as this is merely a repeating device, pushing traffic to all connected nodes, which means that the target machine will still receive non-xPC traffic.

Finally, reduce broadcast traffic on the network as much as possible.