Haichen Shen

MCDNN: An Approximation-Based Execution Framework for Deep Stream Processing Under Resource Constraints

We consider applying computer vision to video on cloud-backed mobile devices using Deep Neural Networks (DNNs). The computational demands of DNNs are high enough that, without careful resource management, such applications strain device battery, wireless data, and cloud cost budgets. We pose the corresponding resource management problem, which we call Approximate Model Scheduling, as one of serving a stream of heterogeneous (i.e., solving multiple classification problems) requests under resource constraints. We present the design and implementation of an optimizing compiler and runtime scheduler to address this problem. Going beyond traditional resource allocators, we allow each request to be served approximately, by systematically trading off DNN classification accuracy for resource use, and remotely, by reasoning about on-device/cloud execution trade-offs. To inform the resource allocator, we characterize how several common DNNs, when subjected to state-of-the art optimizations, trade off accuracy for resource use such as memory, computation, and energy. The heterogeneous streaming setting is a novel one for DNN execution, and we introduce two new and powerful DNN optimizations that exploit it. Using the challenging continuous mobile vision domain as a case study, we show that our techniques yield significant reductions in resource usage and perform effectively over a broad range of operating conditions.

Frame retransmissions considered harmful: improving spectrum efficiency using Micro-ACKs

Retransmissions reduce the efficiency of data communication in wireless networks because of: (i) per-retransmission packet headers, (ii) contention overhead on every retransmission, and (iii) redundant bits in every retransmission. In fact, every retransmission nearly doubles the time to successfully deliver the packet. To improve spectrum efficiency in a lossy environment, we propose a new in-frame retransmission scheme using uACKs. Instead of waiting for the entire transmission to end before sending the ACK, the receiver sends smaller uACKs for every few symbols, on a separate narrow feedback channel. Based on these uACKs, the sender only retransmits the lost symbols after the last data symbol in the frame, thereby adaptively changing the frame size to ensure it is successfully delivered. We have implemented uACK on the Sora platform. Experiments with our prototype validate the feasibility of symbol-level uACK . By significantly reducing the retransmistion overhead, the sender is able to aggressively use higher data rate for a lossy link. Both improve the overall network efficiency. Our experimental results from a controlled environment and an 9-node software radio testbed show that uACK can have up to 140% throughput gain over 802.11g and up to 60% gain over the best known retransmission scheme.

Enable flexible spectrum access with spectrum virtualization

Enabling flexible spectrum access (FSA) in existing wireless networks is challenging due to the limited spectrum programmability - the ability to change spectrum properties of a signal to match an arbitrary frequency allocation. This paper argues that spectrum programmability can be separated from general wireless physical layer (PHY) modulation. Therefore, we can support flexible spectrum programmability by inserting a new spectrum virtualization layer (SVL) directly below traditional wireless PHY, and enable FSA for wireless networks without changing their PHY designs. SVL provides a virtual baseband abstraction to wireless PHY, which is static, contiguous, with a desirable width defined by the PHY. At the sender side, SVL reshapes the modulated baseband signals into waveform that matches the dynamically allocated physical frequency bands - which can be of different width, or non-contiguous - while keeping the modulated information unchanged. At the receiver side, SVL performs the inverse reshaping operation that collects the waveform from each physical band, and reconstructs the original modulated signals for PHY. All these reshaping operations are performed at the signal level and therefore SVL is agnostic and transparent to upper PHY. We have implemented a prototype of SVL on a software radio platform, and tested it with various wireless PHYs. Our experiments show SVL is flexible and effective to support FSA in existing wireless networks.

Accelerating the mobile web with selective offloading

Mobile Web page loads are notoriously slow due to limited computing power and slow network access. Our preliminary experiments show that computation is a significant fraction of page load time on mobile devices. Also, energy arguments suggest that it will stay this way. To compensate the limited computing power, our position is that offloading portions of the page load process to the cloud can significantly improve mobile page load time. We propose a measurement-based framework that allows to offload portions of mobile page load process to the cloud. Unlike browsers that offload fixed parts of page loads such as Opera Mini, our framework will allow to offload any portion of the page load process. We will experiment with a large variety of real-world situations (e.g., varying computing power on mobile devices) by offloading varying portions of page loads using our framework. Informed by the experimental results, we will develop a mobile browser that considers the diverse situations as well as energy and data usage.

Enhancing mobile apps to use sensor hubs without programmer effort

Always-on continuous sensing apps drain the battery quickly because they prevent the main processor from sleeping. Instead, sensor hub hardware, available in many smartphones today, can run continuous sensing at lower power while keeping the main processor idle. However, developers have to divide functionality between the main processor and the sensor hub. We implement MobileHub, a system that automatically rewrites applications to leverage the sensor hub without additional programming effort. MobileHub uses a combination of dynamic taint tracking and machine learning to learn when it is safe to leverage the sensor hub without affecting application semantics. We implement MobileHub in Android and prototype a sensor hub on a 8-bit AVR micro-controller. We experiment with 20 applications from Google Play. Our evaluation shows that MobileHub significantly reduces power consumption for continuous sensing apps.

Fast Video Classification via Adaptive Cascading of Deep Models

Recent advances have enabled oracle classifiers that can classify across many classes and input distributions with high accuracy without retraining. However, these classifiers are relatively heavyweight, so that applying them to classify video is costly. We show that day-to-day video exhibits highly skewed class distributions over the short term, and that these distributions can be classified by much simpler models. We formulate the problem of detecting the short-term skews online and exploiting models based on it as a new sequential decision making problem dubbed the Online Bandit Problem, and present a new algorithm to solve it. When applied to recognizing faces in TV shows and movies, we realize end-to-end classification speedups of 2.4-7.8x/2.6-11.2x (on GPU/CPU) relative to a state-of-the-art convolutional neural network, at competitive accuracy.

MetaSync: Coordinating Storage across Multiple File Synchronization Services

Cloud-based file synchronization services such as Dropbox are a worldwide resource for many millions of users. However, individual services often have tight resource limits, suffer from temporary outages or shutdowns, and sometimes silently corrupt or leak user data. As a solution, the authors design, implement, and evaluate MetaSync, a secure and reliable file synchronization service using multiple cloud synchronization services as untrusted storage providers. To provide global consistency among the storage providers, the authors devise a novel variant of Paxos that enables efficient updates on top of the unmodified APIs exported by each service. MetaSync provides better availability and performance, stronger confidentiality and integrity, and larger storage for end users.

MobileHub: No Programmer Effort for Power Efficiency with Sensor Hub

T oday’s smartphones provide a rich sensing platform that developers have used in tens of thousands of mobile applications. Many of these applications require continuous sensing for tasks ranging from simple step counting to more complex fall detection, sleep apnea diagnoses, dangerous driver monitoring and others. Unfortunately, continuous sensing applications are power-hungry. Interestingly, it is neither the sensors nor the computation that make these applications battery drainers. Instead, the main processor needs to be powered on frequently to collect sensor samples, in turn increasing the power consumption [12, 9, 13]. Hardware manufacturers recognize that supporting low-power continuous sensing is crucial. To this end, companies are embedding a low power microcontroller called a sensor hub in their smartphones [11, 10, 2]. Th e sensor hub continuously collects sensor data keeping the higher power main processor idle. In practice, however, sensor hubs fail to deliver on their power-effi ciency promise. Th e problem is in the diffi culty in programming them. For example, to leverage the sensor hub for a fall detection app, the developer not only needs to write the main application, but also needs to program the sensor hub to sense and notify the main application when a fall is detected. Two approaches have been used to make it easier for developers to program a sensor hub: APIs and hardware SDKs. In the API approach [4, 1], a set of important sensor inference functions are exported via high level APIs to the app developers. Th e problem is that the APIs only support a set of predefi ned events or activities such as step counting. Today, a fall detection application cannot use any of the existing APIs to leverage the sensor hub. It is possible that sensor hub APIs will stabilize, but this is unlikely to happen for many years. Consider how location APIs have evolved since the Java Location API (JSR 179) was introduced in 2003. Sensor hubs themselves have regularly been part of phones since 2011, but it is only in 2014 that a small set of sensor APIs are aligning around common functionality. In the meanwhile, ambitious sensing applications such as BeWell [3] cannot leverage the sensor hub for power effi ciency. In the hardware SDK approach, the developer is provided with specialized tools to directly access the sensor hub. For example, TI provides a proprietary TivaWare Sensor Library [5] to allow developers access to functionality not exposed by soft ware APIs. However, MobileHub: No Programmer Effort for Power Efficiency with Sensor Hub Haichen Shen and David Wetherall University of Washington Aruna Balasubramanian Stony Brook University Anthony LaMarca Intel