In 6.1(1)E3, the IPS uses a different algorithm for managing its idle time, and spends more time polling for packets to reduce latency. This results in CPU usage being reported higher than previous releases,including by external tools such as top and ps. Since the additional CPU load that is reported while polling is actually available to process packets, and reduces as inspection load goes up, it does not negatively affect the overall throughput of the IPS.The statistic reported for Inspection Load is a better indicator of theavailable capacity on the sensor, as it excludes this additional idle-time processing.This is why the cpu uasage is high in ES3.The other reason may be that the main application running on the AIP module could get stuck in a loop, using up most cpu resource.The sensor would stop generating alerts, and the cpu usage would be stuck high on the sensor main application process.The sensor has to be restarted to make it work again.