Non-intrusive load monitoring (NILM) uses electrical measurements taken at a centralized point in a network to monitor many loads downstream. This paper introduces NILM Dashboard, a machine intelligence and graphical platform that uses NILM data for real-time electromechanical system diagnostics. The operation of individual loads is disaggregated using signal processing and presented as time-based load activity and statistical indicators. The software allows multiple NILM devices to be networked together to provide information about loads residing on different electrical branches at the same time. A graphical user interface provides analysis tools for energy scorekeeping, detecting fault conditions, and determining operating state. The NILM Dashboard is demonstrated on the power system data from two United States Coast Guard (USCG) Cutters.

No abstract

Coping with the intermittency of renewable power is a fundamental challenge, with load shifting and grid-scale storage as key responses. We propose Information Batteries (IB), in which energy is stored in the form of information---specifically, the results of completed computational tasks. Information Batteries thus provide storage through speculative load shifting, anticipating computation that will be performed in the future. We take a distributed systems perspective, and evaluate the extent to which an IB storage system can be made practical through augmentation of compiler toolchains, key-value stores, and other important elements in modern hyper-scale compute. In particular, we implement one specific IB prototype by augmenting the Rust compiler to enable transparent function-level precomputation and caching. We evaluate the overheads this imposes, along with macro-level job prediction and power prediction. We also evaluate the space of operation for an IB system, to identify the best case efficiency of any IB system for a given power and compute regime.

It requires significant energy to manufacture and deploy computational devices. Traditional discussions of the energyefficiency of compute measure operational energy, i.e. how many FLOPS in a 50 MW datacenter. However, if we consider the true lifetime energy use of modern devices, the majority actually comes not from runtime use but from manufacture and deployment. In this paper, then, we suggest that perhaps the most climate-impactful action we can take is to extend the service lifetime of existing compute.We design two new metrics to measure how to balance continued service of older devices with the superlinear runtime improvements of newer machines. The first looks at carbon per raw compute, amortized across the operation and manufacture of devices. The second considers use of components beyond compute, such as batteries or radios in smartphone platforms. We use these metrics to redefine device service lifetime in terms of carbon efficiency. We then realize a real-world "junkyard datacenter" made up of Nexus 4 and Nexus 5 phones, which are nearly a decade past their official end-of-life dates. This new-old datacenter is able to nearly match and occasionally exceed modern cloud compute offerings.

No abstract