For provisioning large-scale online applications such as web search, social networks and advertisement systems, data centers face extreme challenges in providing low latency for short flows (that result from end-user actions) and high throughput for background flows (that are needed to maintain data consistency and structure across massively distributed systems). We propose L 2 DCT, a practical data center transport protocol that targets a reduction in flow completion times for short flows by approximating the Least Attained Service (LAS) scheduling discipline, without requiring any changes in application software or router hardware, and without adversely affecting the long flows. L 2 DCT can co-exist with TCP and works by adapting flow rates to the extent of network congestion inferred via Explicit Congestion Notification (ECN) marking, a feature widely supported by the installed router base. Though L 2 DCT is deadline unaware, our results indicate that, for typical data center traffic patterns and deadlines and over a wide range of traffic load, its deadline miss rate is consistently smaller compared to existing deadlinedriven data center transport protocols. L 2 DCT reduces the mean flow completion time by up to 50% over DCTCP and by up to 95% over TCP. In addition, it reduces the completion for 99th percentile flows by 37% over DCTCP. We present the design and analysis of L 2 DCT, evaluate its performance, and discuss an implementation built upon standard Linux protocol stack.

Abstract-Autonomous underwater vehicles (AUVs) are indispensable tools for marine scientists to study the world's oceans. Depending on their missions, AUVs are equipped with advanced sensors (sonar, cameras, acoustic communication, bio-sensors), have on-board computers for data analysis (image analysis, data compression), and are capable of on-board decision making (resource planning, swarming). Since AUVs operate solely on battery power, power and energy management is a crucial issue. Missioncritical tradeoff decisions have to be made between energy consumption and sensing, data processing, and communication activities. Mission planning has to consider these tradeoffs when provisioning resources for expected future events, or when dealing with changing environmental conditions such weather, water currents, and seafloor profiles. Effective power and energy management requires knowledge about the actual energy consumption of each active component within the AUV. Effective planning requires simulators that can predict energy consumptions based on expected future events and environmental conditions.In this paper, we discuss the design and implementation of a power measurement infrastructure for the Teledyne Webb research Slocum glider. This infrastructure can be used for online power/energy management or to better understand the time-dependent energy consumption profile of the active glider components during a particular mission. We also discuss the design of a new simulation environment for the Slocum glider which uses the power/energy data obtained by our measurement infrastructure, in addition to seafloor and coastal radar information. We illustrate the effectiveness of the new tools in the context of planning a glider flight across the continental shelf off the coast of New Jersey.

No abstract

No abstract

We propose Style Conditioned Recommendations (SCR) and introduce style injection as a method to diversify recommendations. We use Conditional Variational Autoencoder (CVAE) architecture, where both the encoder and decoder are conditioned on a user profile learned from item content data. This allows us to apply style transfer methodologies to the task of recommendations, which we refer to as injection. To enable style injection, user profiles are learned to be interpretable such that they express users' propensities for specific predefined styles. These are learned via label-propagation from a dataset of item content, with limited labeled points. To perform injection, the condition on the encoder is learned while the condition on the decoder is selected per explicit feedback. Explicit feedback can be taken either from a user's response to a style or interest quiz, or from item ratings. In the absence of explicit feedback, the condition at the encoder is applied to the decoder. We show a 12% improvement on N DCG@20 over the traditional VAE based approach and an average 22% improvement on AUC across all classes for predicting user style profiles against our best performing baseline. After injecting styles we compare the user style profile to the style of the recommendations and show that injected styles have an average +133% increase in presence. Our results show that style injection is a powerful method to diversify recommendations while maintaining personal relevance. Our main contribution is an application of a semi-supervised approach that extends item labels to interpretable user profiles.