Project Description

Different stakeholders, such as end users, networking researchers, network operators, policymakers, etc., rely on Internet measurement tools to assess network quality. However, despite significant progress in developing new methodologies to measure and analyze Internet quality, existing tools fail to answer basic questions. Users often have a mismatch in reporting from state-of-the-art network measurement tools and their network experience.  For example, when the speed test measurements report ideal network conditions, users might still experience poor quality, e.g., excessive rebuffering events for video streaming sessions, high page load times for web browsing sessions, or low resolution for video conferencing sessions, etc. We can attribute this mismatch to the quality of the data collected by existing Internet measurement tools.

More specifically, most of the existing network measurement tools embrace a simplified design and have the following attributes:
      (1) Stateless, i.e., they only collect measurement data once, and their data collection is not driven by the observations from past measurements;
      (2) Unimodal, i.e., they only collect active or passive data, not both; and
      (3) Single-view, i.e., they only use a single vantage point (e.g., client's end device, cable modem, etc.) for data collection.

Whilst a simplified design may attract a larger user base and make participation easier for end-users, our previous work has shown that it can also result in the collection of low-quality data that lacks crucial measurement context. This poses challenges in accurately assessing the overall state of the network and providing meaningful feedback to stakeholders.

This project explores the development of a new network measurement tool, NetFlex, that improves data quality while minimizing the participation threshold for end-users. It complements existing network monitoring tools/infrastructure with low-cost single-board computers (SBCs), deployed across the UCSB campus, to enable iterative active and passive network measurement data collection from one or more vantage points. The interactive user interface of our system provides users with the most relevant information at each step while also utilizing user inputs to shape future network measurement objectives. The system translates high-level user intents into target-specific commands and configurations, allowing data to be collected from multiple vantage points, such as end hosts, gateway routers, and measurement servers. Once data is collected, it is transformed into universal network data representations to facilitate easy ingestion by various data-processing methodologies in a privacy-preserving manner. The collected data can then be analyzed by different analytics modules to extract meaningful insights. Finally, the system presents the insights to the end-users for feedback.

 

Team Members

  • Akul Singh
  • Evania Cheng
  • Karthik Bhattaram
  • Natchanon (Ken) Thampiratwong

 

Professor and Mentors

  • Prof. Arpit Gupta
  • Grad mentors: Sylee Beltiukov and Jaber Daneshamooz