This will be due to some device flooding your trap listener with traps. You need to track down the culprit device and either fix the issue on the device or change the trap level on the device
Thank you Capt. Obvious,
SQL is installed in a dedicated machine; but we have noticed that everytime we launch a maintenance of the database, the cpu load decreases down to 20% for 2 or 3 days and then comes back to 100%.
We are still looking for the culprit device in order to fix the issue...