Client-side video players employ adaptive bitrate (ABR) algorithms to optimize user quality of experience (QoE). Despite the abundance recently proposed schemes, state-of-the-art ABR suffer from a key limitation: they use fixed control rules based on simplified or inaccurate models deployment environment. As result, existing schemes inevitably fail achieve optimal performance across broad set network conditions and QoE objectives.

Over the past two or three years, wireless cellular networks have become faster than before, most notably due to deployment of LTE, HSPA+, and other similar networks. LTE throughputs can reach many megabits per second even rival WiFi in some locations. This paper addresses a fundamental question confronting transport application-layer protocol designers: which network should an application use? WiFi, Multi-Path TCP (MPTCP) running over both?

To cope with the high resource (network and compute) demands of real-time video analytics pipelines, recent systems have relied on frame filtering. However, filtering has typically been done neural networks running edge/backend servers that are expensive to operate. This paper investigates on-camera filtering, which moves beginning pipeline. Unfortunately, we find commodity cameras limited compute resources only permit via differencing based low-level features. Used incorrectly, such...

This demo presents a measurement toolkit, Mahimahi, that records websites and replays them under emulated network conditions. Mahimahi is structured as set of arbitrarily composable UNIX shells. It includes two shells to record replay Web pages, RecordShell ReplayShell, well for emulation, DelayShell LinkShell. In addition, corpus recorded along with benchmark results link traces (https://github.com/ravinet/sites). improves on prior record-and-replay frameworks in three ways. First, it...

The existing slowness of the web on mobile devices frustrates users and hurts revenue website providers. Prior studies have attributed high page load times to dependencies within process: network latency in fetching a resource delays its processing, which turn when dependent resources can be discovered fetched.

Video-conferencing applications impose high loads and stringent performance requirements on the network. To better understand manage these applications, we need effective ways to measure in wild. For example, measurements would help network operators capacity planning, troubleshooting, setting QoS policies. Unfortunately, large-scale of production networks cannot rely end-host cooperation, an in-depth analysis packet traces requires knowledge header formats. Zoom is one most sophisticated...

This demo presents a measurement toolkit, Mahimahi, that records websites and replays them under emulated network conditions. Mahimahi is structured as set of arbitrarily composable UNIX shells. It includes two shells to record replay Web pages, RecordShell ReplayShell, well for emulation, DelayShell LinkShell. In addition, corpus recorded along with benchmark results link traces (https://github.com/ravinet/sites).

A graph neural network (GNN) enables deep learning on structured data. There are two major GNN training obstacles: 1) it relies high-end servers with many GPUs which expensive to purchase and maintain, 2) limited memory cannot scale today's billion-edge graphs. This paper presents Dorylus: a distributed system for GNNs. Uniquely, Dorylus can take advantage of serverless computing increase scalability at low cost. The key insight guiding our design is computation separation. Computation...

We propose Accel-Brake Control (ABC), a simple and deployable explicit congestion control protocol for network paths with time-varying wireless links. ABC routers mark each packet an "accelerate" or "brake", which causes senders to slightly increase decrease their windows. Routers use this feedback quickly guide towards desired target rate. requires no changes header formats user devices, but achieves better performance than XCP. is also incrementally deployable; it operates correctly when...

Remote dependency resolution (RDR) is a proxy-driven scheme for reducing mobile page load times; proxy loads requested using local browser, fetching the page's resources over fast proxy-origin links instead of client's slow last-mile links. In this paper, we describe two fundamental challenges to efficient RDR proxying: increasing popularity encrypted HTTPS content, and fact that, due time-dependent network conditions properties, proxying can actually increase times. We solve these problems...

Users in developing regions still suffer from poor web performance, mainly due to their unique landscape of low-end devices. In this paper, we uncover a root cause suboptimal performance by cross-analyzing longitudinal resource (in particular, memory) profiles large social network, and the memory consumption modern webpages five regions. We discover that primary culprit for hitting constraints is JavaScript execution which existing optimizations are ill-suited alleviate. To handle this,...

Motivated by the rapid emergence of programmable switches, network interface cards, and software packet processing, this paper asks: given a task (e.g., virtualization or measurement) in network, should we implement it at network's end hosts (the edge) its switches core)? To answer question, analyze range common tasks spanning virtualization, deep inspection, measurement, application acceleration, resource management. We conclude that, while edge is better even required for certain access...

Machine learning inference platforms continue to face high request rates and strict latency constraints. Existing solutions largely focus on compressing models substantially lower compute costs (and time) with mild accuracy degradations. This paper explores an alternate (but complementary) technique that trades off resource a perinput granularity: early exit models, which selectively allow certain inputs model from intermediate layer. Though intuitive, exits fundamental deployment...

Providers of location-based services seek new methods to authenticate the location their clients. We propose a novel infrastructure-based solution that provides spontaneous and transaction-oriented mobile device authentication via an integrated 802.11× wireless access point 3G femtocell system. By simply making voice call while remotely monitoring activity, calling party can verify (co-operating) called party's even when participants have no pre-existing relationship. show how such traffic...

Electro-optic (EO) image sensors exhibit the properties of high resolution and low noise level at daytime, but they do not work in dark environments. Infrared (IR) poor cannot separate objects with similar temperature. Therefore, we propose a novel framework IR enhancement based on information (e.g., edge) from EO images, which improves images helps us distinguish night. Our superimposing/blending edges onto corresponding transformed their resolution. In this framework, adopt theoretical...

Mobile browsers suffer from unnecessary cache misses. The same binary object is often named by multiple URLs which correspond to different keys. Furthermore, servers frequently mark objects as uncacheable, even though the objects' content stable over time.

This paper makes the case for "Room-Area Networks" (RAN), a new category that falls between personal area networks and local networks. In RAN, set of nodes can hear each other only if they are in same room, broadly construed as being within earshot. We define RAN abstraction, we present example applications ranging from social contact management to building automation gaming where this abstraction will help. The requirements poorly served by current technologies such Bluetooth, near-field...

Spontaneous and robust mobile device location authentication can be realized by supplementing existing 802.11x access points (AP) with small cells. We show that transferring network traffic to a computing associated femtocell while remotely monitoring its ingress activity, any internet-connected sender verify the cooperating receiver's location. describe prototype non-cryptographic system we constructed, explain how design both voice data transmissions distinct, discernible signatures. Using...

DNN models across many domains continue to grow in size, resulting high resource requirements for effective training, and unpalatable (and often unaffordable) costs organizations research labs scales. This paper aims significantly reduce training with use of preemptible instances, i.e., those that can be obtained at a much cheaper price while idle, but may preempted whenever requested by priority users. Doing so, however, requires new forms resiliency efficiency cope the possibility frequent...