Student Projects

Contact Person: Taveesh Sharma, Tarun Mangla

Goal: Add more measurements to the Netrics suite including:

Speed test Diagnostics: Was the bottleneck link during a speed test the ISP link? This is important given speed test data is often used to assess ISP performance. From a user’s perspective, this additional information can help them contextualize the results. One way to assess that is by sending probes to the access link while the test is running. If the access link is congested, it is likely that we will observe inflated delays (due to queuing). We first need to design the methodology and validate it in the lab. After validation, we can integrate with netrics and characterize for how many tests, the ISP link was the bottleneck. There is also potential to integrate the test with NDT7, although only for their command-line tool.

Application-layer measurements: Run Netflix- and Zoom-like videos on Netrics devices and log the application performance. We have a beta version of the VCA test — based on WebRTC. We can also reach out to the UCSB folks. These tests could be useful for the QoE inference project.

Lightweight Speedtest: Existing speed test consume a significant amount of data especially at high speeds. Given the data caps ISPs impose, this limits the ability to take frequent speed measurements. Past work has proposed UDP-based methods that rely on inter-arrival times to infer speed. A more recent work proposes TCP with modified congestion control for lightweight speed testing. A starting step can be to evaluate the accuracy of these techniques to test speed.

Skills: Python; Linux Command Line; Networking; Network Measurements

Contact Person: Jesse London, Marc Richardson

Goal: Help update the Netrics framework and port existing tests to the new version.

Contribute to our open-source software for continuous, fine-grained measurement of Internet performance.

Skills: Basic Networking; Python; Git

Contact Person: Tarun Mangla

Goal: Assist the Chicago Area Broadband Initiative (CABI) evaluate the performance of a wireless-based network. The primary goal is to measure wireless-based networks. There are two networks of interest. One of them uses a 60GHz Siklu backhaul and the other network uses a 70GHz wireless backhaul. The networks will be used to serve internet to residents in public housing. The goal is to deploy Netrics devices that measure the performance of the network over time. In addition, we will collect radio-level logs from the antenna themselves and the weather data. The Netrics performance data then can be correlated with both radio-level and weather logs.

Skills: Wireless networking

Contact Person: Nick Feamster, Nicole Marwell, Tarun Mangla

Goal: Use the existing Netrics data to compare various metrics of Internet performance across neighborhoods. This analysis will build on existing analysis of the differences in throughput between the Chicago neighborhoods of South Shore and Logan Square. The project will focus on extending the analysis using the latency data. There is also a possibility of trying other kinds of analysis such as using hierarchical models.

Skills: Basic statistics; Familiar with data science libraries in Python; Basic visualization

Contact Person: Marc Richardson, Tarun Mangla

Goal: Participants are asked to refrain from moving or changing their Internet service while they are enrolled in the Internet study and have a device installed on their access network. This eligibility criterion is to ensure that data from devices can be assumed to capture Internet performance for one access network over time. Moving or changing Internet service creates discrepancies in the data that inhibit statistic inference and valid data analysis.

Despite this requirement of our study, we still have some participants who are moving or changing ISPs during the study (i.e., non-compliance). An analysis of the measurement data collected to date can help identify participants who might have switched ISPs or moved, which can be useful for contextualizing and annotating the data we release to the public. The final deliverable for this project is a process for inferring events like moving and switching internet service from the measurement data as well as data annotation to flag such events in our measurement datasets.

Skills: Python; Anomaly detection; Data processing

Contact Person: Marc Richardson

Goal: The Internet Equity Initiative collects survey responses from study participants. The survey provides information about the participants’ internet service (provider, speed tier, cost), internet usage, internet access equipment, and demographics. Survey data can be linked to measurement data that we collect from the study devices through a device ID field that is included in both the measurement datasets and the survey dataset.

Before we can begin to analyze the measurement and survey data jointly, the survey data must be cleaned and processed. For example, some of the data collected through the survey are not machine-readable and need to be encoded to facilitate analysis. In addition, to prepare the survey data for publication, PII needs to be stripped out of the dataset. A preliminary analysis of the survey responses is also needed to determine the quality of data collected through the survey and identify ineffective questions or missing questions. The final deliverable for this project is 1) a cleaned survey dataset that is ready to be linked to measurement data for analysis and 2) a report with insights and recommendations about the quality of the data and the survey design.

Skills: Python; Data Cleaning; Survey Design

Contact Person: Tarun Mangla

Goal: The project considers if permit data can be used to map wired and wireless infrastructure. Given that a lot of infrastructure in the U.S. is built over the public right-of-way, cellular infrastructure providers usually need to take permissions from the local city governments to occupy and/or dig the public ROW. These permits are recorded by city governments for logging purposes and also in some cases made public to increase transparency in governance. Therefore, the underlying question is what is the feasibilty of using permit data to map last-mile cellular network radio sites?

Skills: Python, Data Analytics