protocol / research

Research at Protocol Labs
220 stars 20 forks source link

MobiProbe: A Large-Scale Mobile Measurement Platform #25

Closed zalqudah closed 2 years ago

zalqudah commented 3 years ago

MobiProbe: A Large-Scale Mobile Measurement Platform

Description

The Internet is constantly evolving. The availability of high-quality Internet measurements is a key to understanding the Internet performance as well as to finding ways to address potential sub-optimalities. Over the past few decades, several measurement platforms were created to allow researchers and network operators to study the Internet and monitor services. For instance, PlatnetLab has achieved a significant success. It has been used by massive number of Internet measurement studies that have improved our understanding of the Internet. PlanetLab nodes (vantage points) were mostly servers connected to a high-capacity networks on academic institutions and other companies. Ripe Atlas is another measurement platform that employed hardware probes (and recently software probes) that are hosted on residential and institutional networks. This platform continues to achieve a significant success enabling network operators to run measurement experiments to monitor and troubleshoot their networked services. Also, a significant Internet research literature has been created with the help of this platform.

Measurement studies always aim at creating a measurement setup that is highly representative of the Internet. Today's Internet services are mostly accessed via mobile devices. While there have been numerous research proposals that attempted to create a mobile measurement platform, these projects remain very limited in terms of deployment success. The purpose of this project is to create and operate a platform that complements the existing Ripe Atlas platform by providing probes representing mobile devices ( and perhaps IoT devices). We do not plan to compete with the existing proposals, but to learn from and complement them to create and operate a large-scale mobile measurement platform. Such a platform will have tremendous impact on our understanding of mobile networks and will nicely complement the existing Ripe Atlas platform.

State of the Art

Ripe Atlas is a prominent Internet measurement platform that is widely available to network operators and researchers. The platform is a credit-based testbed where vantage points are hardware devices that perform network measurements. Recently the platform has developed software probes to expand the measurement network.

Several research projects proposed various mobile measurement platforms. For instance, MobiPerf is a measurement app that was developed at University of Michigan. It supports basic measurements such as ping, traceroute, and TCP throughput. Unfortunately, this app was last updated in 2014. Mobile Internet Testbed for Application Traffic Experimentation (MITATE) is a platform that was developed at Montana State University. Per the project website, it has been frozen as of 2017 due to students graduating.

Today, we do not have a widely available mobile measurement platform that can be used to study a significant part of the Internet (mobile devices and networks).

Solving this Open Problem

Due to the importance of a mobile measurement platform, we intend to focus on building and operating a mobile measurement platform. The design of this platform will learn from existing proposals. We will primarily focus on building a high-quality platform and maintaining it. A critical issue in this regard is recruiting hosts that are willing to act as a mobile vantage points. We plan to survey various stake-holders (e.g., network operators, researchers, and the public) to find out what could motivate them to join this measurement networks. Based on the result of this survey we can design a scheme to ensure enough recruits for a meaningful and representative measurement platform.

Impact

zalqudah commented 3 years ago

@miyazono I'd love your thoughts.

yiannisbot commented 2 years ago

Great topic and indeed a valuable contribution to the research community. We have not been able to fund projects in this space so far, hence, the delay in getting back to you. However, we have recently started a more focused effort to measure Web3 networking platforms, and in particular we've started from IPFS and libp2p. We're focusing on the fixed part of the network, rather than the mobile. This is not because the mobile is less important, but just because this is where our networks currently operate and this is where we need better visibility into protocol performance.

If you have interest and work in similar topics and would like to get involved, please write your thoughts and ways you'd imagine your measurement and monitoring methodologies would be impactful in a new issue. Thanks!