2-Step Power Scheduling with Adaptive Control Interval for Network Intrusion Detection Systems on Multicores

Network intrusion detection system (NIDS) is becoming an important element even in embedded systems as well as in data centers since embedded computers have been increasingly exposed to the Internet. The demand for power budget of these embedded systems is a critical issue in addition to that for performance. In this paper, we propose a technique to minimize power consumption in the NIDS called by 2-step power scheduling with the adaptive control interval. In addition, we also propose a CPU-core controlling algorithm so that our scheduling technique can preserve the performance for other applications and NIDS assuming the cases of multiplexing NIDS and them simultaneously on the same device such as a home server or a mobile platform. We implement our 2-step algorithm into Suricata, which is a popular NIDS, as well as a 1-step algorithm and a simple fixed interval algorithm for evaluations. Experimental results show that our 2-step scheduling with both the adaptive and the fixed 30-millisecond interval achieve 75% power saving comparing with the Ondemand governor and 87% comparing with the Performance governor in Linux, respectively, without affecting their performance capability on four ARM Cortex-A15 cores at the network traffic of 1,000 packets/seconds. In contrast, when the network traffic reaches to 17,000 packets/seconds, our 2-step scheduling and the Ondemand as well as the Performance governor can maintain the packet processing capacity while the fixed 30-milliseconds interval processes only 50% packets with two and three cores, and about 80% packets on four cores.