Analyzing survivability in a Power over Ethernet system


Anne Remke


6th Sept. 2007






Power over Ethernet (PoE) transmits electrical power, along with data, over standard twisted-pair cable in an Ethernet network. Together with un-interruptible power supply (UPS) units this can be used to provide seamless and reliable power over Ethernet cabling. We are planning to analyze a setting where several surveillance devices are powered over Ethernet cabling. In case commercial power fails a battery storage provides power as long as possible. When the battery is empty before commercial power is restored the surveillance devices do not receive any power and need to shut down. This is considered as a not recoverable calamity/disaster. The overall scenario is build as "given-occurrence-of-disaster" (GOOD) model and modeled as Continuous Time Markov Chain (CTMC). This allows us to use existing CSL specifications for survivability to express measures of interest. To be able to dimension the size of the UPS and to calculate the rate that leads to calamity, once the system is in the so-called "battery mode" we need to model the system and its workload in more detail. Each surveillance device is modeled with two queues, one for the incoming packet stream and one for the outgoing packet stream, leading to an open queueing network. Based on our recent work on model checking Labeled Jackson Queueing networks our goal is to compute the power consumption for a given workload model and thus to set the calamity rate in the survivability model. This hierarchical modeling approach provides a lot of modeling flexibility. The general survivability model can be used for any type of service center, only the underlying queueing network needs to be adapted to calculate the calamity rate.