Cloud computing and AI workloads are driving unprecedented demand for efficient communication within and across datacenters. However, the coexistence of intra- and inter-datacenter traffic within datacenters plus the disparity between the RTTs of intra- and inter-datacenter networks complicates congestion management and traffic routing. Particularly, faster congestion responses of intra-datacenter traffic causes rate unfairness when competing with slower inter-datacenter flows. Additionally, inter-datacenter messages suffer from slow loss recovery and, thus, require reliability. Existing solutions overlook these challenges and handle inter- and intra-datacenter congestion with separate control loops or at different granularities. We propose Uno, a unified system for both inter- and intra-DC environments that integrates a transport protocol for rapid congestion reaction and fair rate control with a load balancing scheme that combines erasure coding and adaptive routing. Our findings show that Uno significantly improves the completion times of both inter- and intra-DC flows compared to state-of-the-art methods such as Gemini.

Modular data centers (MDCs) that can be placed right at the energy farms and powered mostly by renewable energy, is a flexible and effective approach to lowering the carbon footprint of data centers. However, the main challenge of using renewable energy is the high variability of power produced, which implies large volatility in powering computing resources at MDCs, and degraded application performance due to the task evictions and migrations. This causes challenges for platform operators to decide the MDC deployment.

Heuristics are ubiquitous in computer systems. Examples include congestion control, adaptive bit rate streaming, scheduling, load balancing, and caching. In some domains, theoretical proofs have provided clarity on the conditions where a heuristic is guaranteed to work well. This has not been possible in all domains because proving such guarantees can involve combinatorial reasoning making it hard, cumbersome and error-prone. In this paper we argue that computers should help humans with the combinatorial part of reasoning. We model reasoning questions as ∃∀ formulas [1] and solve them using the counterexample guided inductive synthesis (CEGIS) framework. As preliminary evidence, we prototype CCmatic, a tool that semi-automatically synthesizes congestion control algorithms that are provably robust. It rediscovered a recent congestion control algorithm that provably achieves high utilization and bounded delay under a challenging network model. It also found previously unknown variants of the algorithm that achieve different throughput-delay trade-offs.

Renewable energy is becoming an important power source for data centers, especially with the zero-carbon waste pledges made by big cloud providers. However, one of the main challenges of renewable energy sources is the high variability of power produced. Traditional approaches such as batteries or transmitting to the grid fall short on scale, overhead, or "green-ness". We propose Virtual Battery: instead of adapting the availability of power to match the computation demand we shift computational demand to meet the availability of power. Virtual batteries shift demand by requiring applications to either be flexible and delay-tolerant or proactively migrating to where power is (going to be) available. We show that using multiple virtual battery sites in combination can meet the needs of modern applications. Moreover, we show how an intelligent network and power aware co-scheduler can not only provide availability despite variability but also help mitigate migration related network overhead by over 30% in total and 4.2x at peak.

Sketching algorithms or sketches have emerged as a promising alternative to the traditional packet sampling-based network telemetry solutions. At a high level, they are attractive because of their high resource efficiency and provable accuracy guarantees. While there have been significant recent advances in various aspects of sketching for networking tasks, many fundamental challenges remain unsolved that are likely stumbling blocks for adoption. Our contribution in this paper is in identifying and formulating these research challenges across the ecosystem encompassing network operators, platform vendors/developers, and algorithm designers. We hope that these serve as a necessary fillip for the community to enable the broader adoption of sketch-based telemetry.

In 2017, smart speakers (such as Amazon Echo, Google Home, etc.) became a commercial success. Most smart speakers have a circular microphone array to provide hands-free, voice-only interaction from a distance. In this work, we exploit this mic array for opportunistically sensing gestures and tracking exercises. To this end, we measure the Doppler shift on a pilot tone caused by a gesturing human body, and use beamforming of the mic array to extend the range of the detection. Data from 12 participants show that gestures can be detected with an accuracy of 96.8% up to a distance of 2.5 meters using an inaudible 20 kHz pilot tone. For exercise tracking, we train a deep neural network to recognize 10 different exercises, and count repetitions by peak-finding heuristics. Data from 17 participants show that exercise classification accuracy is 96% and count accuracy is 91.8%. To conclude, we discuss hardware enhancements to smart speakers to further increase their gesture sensing capabilities.

Application-specific ICs have been traditionally used to support the high computational and data rate requirements in medical ultrasound systems, particularly in receive beamforming. Utilizing the previously developed efficient front-end algorithms, in this paper, we present a simple programmable computing architecture, consisting of a field-programmable gate array (FPGA) and a digital signal processor (DSP), to support core ultrasound signal processing. It was found that 97.3% and 51.8% of the FPGA and DSP resources are, respectively, needed to support all the front-end and back-end processing for B-mode imaging with 64 channels and 120 scanlines per frame at 30 frames/s. These results indicate that this programmable architecture can meet the requirements of low- and medium-level ultrasound machines while providing a flexible platform for supporting the development and deployment of new algorithms and emerging clinical applications.

Over the last decade, a picture archiving and communications system (PACS) has been accepted by an increasing number of clinical organizations. Today, PACS is considered as an essential image management and productivity enhancement tool. Image compression could further increase the attractiveness of PACS by reducing the time and cost in image transmission and storage as long as 1) image quality is not degraded and 2) compression and decompression can be done fast and inexpensively. Compared to JPEG, JPEG 2000 is a new image compression standard that has been designed to provide improved image quality at the expense of increased computation. Typically, the decompression time has a direct impact on the overall response time taken to display images after they are requested by the radiologist or referring clinician. In this paper, we present a fast JPEG 2000 decoder running on a low-cost programmable processor. It can decode a losslessly compressed 2048 x 2048 CR image in 1.51 s. Using this kind of a decoder, performing JPEG 2000 decompression at the PACS display workstation right before images are displayed becomes viable. A response time of 2 s can be met with an effective transmission throughput between the central short-term archive and the workstation of 4.48 Mb/s in case of CT studies and 20.2 Mb/s for CR studies. We have found that JPEG 2000 decompression at the workstation is advantageous in that the desired response time can be obtained with slower communication channels compared to transmission of uncompressed images.