Electrical Engineering & Computer Sciences (EECS)

The edge fringing effect (EFE) on the capacitance-voltage (C -V) responses from depletion to deep depletion (DD) of p-substrate metal-oxide-semiconductor capacitors with ultrathin oxide (2.1-2.9 nm) and high-κ dielectric was examined. It was found that the EFE plays a significant role on the initiation voltage of DD in the C -V curve. The shape of capacitance in DD has been characterized in detail via a modified quantum-mechanical-based EFE model with the consideration of edge direct tunneling current conduction concept. Excellent agreement between the model and the experimental data for gate oxide has been achieved. To elucidate the enhanced fringing effect at edge, the effective surface field ratio between edge and bulk was adopted to clarify the role of injected carriers with the EFE mechanism. It was found that the extracted ratio increases with the oxide thickness and the initiation voltage of DD accordingly.

This paper presents a predictive controller for handling plug-and-play (P&P) charging requests of electric vehicles (EVs) in a distribution system. The proposed method uses a two-stage hierarchical control scheme based on a model predictive control (MPC) formulation for tracking periodic references. The first stage computes a reachable periodic reference that trades off deviation from the nominal voltage with the required generation control. The second stage computes a controller that tracks this reference and charges the EVs, while satisfying system constraints at all times. Under the assumption of a time-periodic load, it is shown that the proposed controller is recursively feasible and exponentially stable i.e. the EVs' state of charge (SOC) and bus voltages converge to the desired SOC and to the optimal periodic reference respectively. Finally, the proposed scheme is illustrated in a set of examples.

Spark started out of our research group's discussions with Hadoop users at and outside UC Berkeley. We saw that as organizations began loading more data into Hadoop, they quickly wanted to run rich applications that the single-pass, batch processing model of MapReduce does not support efficiently. In particular, users wanted to run: u More complex, multi-pass algorithms, such as the iterative algorithms that are common in machine learning and graph processing u More interactive ad hoc queries to explore the data

Abstract Debugging the massive parallel computations that run in today's datacenters is hard, as they consist of thousands of tasks processing terabytes of data. It is especially hard in production settings, where performance overheads of more than a few percent are unacceptable. To address this challenge, we present Arthur, a new debugger that provides a rich set of analysis tools at close to zero runtime overhead through selective replay of data flow applications.