Search for: All records

Optical fiber deployments in metropolitan areas are critical for information distribution to businesses and large segments of the population. In this paper, we describe a char- acterization study of metropolitan area fiber networks in the US. The goal of our work is to elucidate the key aspects of these infrastructures and to assess how they can be enhanced to support growth in cloud-mobile via expanded connectivity to data centers. We collect maps of 204 metro fiber networks and transcribe these into a geographic information system for analysis and visualization. We report on characteristics including raw miles, geography, proximity to users, correspondence to other infrastructure and PoP/data center proximity. These characteris- tics indicate highly diverse deployments in different metro areas and suggest different strategies for future deployments. Next, we conduct a resource allocation analysis to assess how fiber infrastructure can be deployed in metro areas to reduce the physical distance to data centers over a range of cost scenarios. Our results show that a small number of new connections to data centers can significantly reduce physical distances to users. 

In this paper, we report on our investigation of how current local time is reported accurately by devices connected to the internet. We describe the basic mechanisms for time management and focus on a critical but unstudied aspect of managing time on connected devices: the time zone database (TZDB). Our longitudinal analysis of the TZDB highlights how internet time has been managed by a loose confederation of contributors over the past 25 years. We drill down on details of the update process, update types and frequency, and anomalies related to TZDB updates. We find that 76% of TZDB updates include changes to the Daylight Saving Time (DST) rules, indicating that DST has a significant influence on internet-based time keeping. We also find that about 20% of updates were published within 15 days or less from the date of effect, indicating the potential for instability in the system. We also consider the security aspects of time management and identify potential vulnerabilities. We conclude with a set of proposals for enhancing TZDB management and reducing vulnerabilities in the system. 

In this paper, we describe an architecture for clock synchronization in IoT devices that is designed to be scalable, flexibly accommodate diverse hardware, and maintain tight synchronization over a range of operating conditions. We begin by examining clock drift on two standard IoT prototyping platforms. We observe clock drift on the order of seconds over relatively short time periods, as well as poor clock rate stability, each of which make standard synchronization protocols ineffective. To address this problem, we develop a synchronization system, which includes a lightweight client, a new packet exchange protocol called SPoT and a scalable reference server. We evaluate the efficacy of our system over a range of configurations, operating conditions and target platforms. We find that SPoT performs synchronization 22x and 17x more accurately than MQTT and SNTP, respectively, at high noise levels, and maintains a clock accuracy of within ∼15ms at various noise levels. Finally, we report on the scalability of our server implementation through microbenchmark and wide area experiments, which show that our system can scale to support large numbers of clients efficiently. 

One-way delay (OWD) between end hosts has important implications for Internet applications, protocols, and measurement-based analyses. We describe a new approach for identifying OWDs via passive measurement of Network Time Protocol (NTP) traffic. NTP traffic offers the opportunity to measure OWDs accurately and continuously from hosts throughout the Internet. Based on detailed examination of NTP implementations and in-situ behavior, we develop an analysis tool that we call TimeWeaver, which enables assessment of precision and accuracy of OWD measurements from NTP. We apply TimeWeaver to a ∼1TB corpus of NTP traffic collected from 19 servers located in the US and report on the characteristics of hosts and their associated OWDs, which we classify in a precision/accuracy hierarchy. To demonstrate the utility of these measurements, we apply iterative hard-threshold singular value decomposition to estimate the missing OWDs between arbitrary hosts from the highest tier in the hierarchy. We show that this approach results in highly accurate estimates of missing OWDs, with average error rates on the order of less than 2%. 

Understanding the nature and characteristics of Internet events such as route changes and outages can serve as the starting point for improvements in network configurations, management and monitoring practices. However, the scale, diversity, and dynamics of network infrastructure makes event detection and analysis challenging. In this paper, we describe a new approach to Internet event measurement, identification and analysis that provides a broad and detailed perspective without the need for new or dedicated infrastructure or additional network traffic. Our approach is based on analyzing data that is readily available from Network Time Protocol (NTP) servers. NTP is one of the few on-by-default services on clients, thus NTP servers have a broad perspective on Internet behavior. We develop a tool for analyzing NTP traces called Tezzeract, which applies Robust Principal Components Analysis to detect Internet events. We demonstrate Tezzeract’s efficacy by conducting controlled experiments and by applying it to data collected over a period of 3 months from 19 NTP servers. We also compare and contrast Tezzeract’s perspective with reported outages and events identified through active probing. We find that while there is commonality across methods, NTP-based monitoring provides a unique perspective that complements prior methods.