Lead PIs

• Z. Morley Mao, University of Michigan
• David Choffnes

Abstract

Despite over a decade of Internet access from smartphones, we still have little understanding of the network performance we receive. A number of small testbeds and user studies have enabled progress in the face of these challenges, but with limited scope, duration, coverage and generality.

This previous work suffers from three key limitations that hamper their success.

• First, these individual solutions do not scale: each individual app or measurement platform is inherently limited to the population of participating users running a single piece of software.
• Second, each solution is inconsistent and inflexible in the set of network measurements it supports and the contextual information describing the experimental environment, making it difficult to ensure scientific rigor and to merge disparate datasets.
• Third, these solutions are uncoordinated in how they conduct network measurements: multiple apps can wastefully measure the same property independently or, worse, interfere with each other by running measurements at the same time from the same device.

Mobilyzer addresses these issues by providing a scalable, efficient and controllable open platform for network measurement from mobile devices. Mobilyzer provides the following components:

• Mobile measurement library for apps. An app-based deployment model is uniquely capable of capturing the detailed view of mobile systems described above. Our Mobilyzer library provides standard measurement tools for users, researchers and developers, and it manages data collection by reporting to cloud-based servers.
• Measurement manager. Mobilyzer allows researchers to conduct mobile network experiments using a collection of mobile devices running apps. To support a wide range of existing and future experiments, the measurement manager can assign programmable collections of measurements to devices in a way that optimizes for efficient use of the available, limited resources. In addition to scheduling measurements, this system coordinates across multiple devices to ensure that measurements do not overwhelm any particular device, network or host being probed.
• Data collection, analysis and archival. We use a cloud-based system to collect, analyze, anonymize and publish data reported from mobile devices. This approach simplifies data management and sharing, provides a centralized repository for tools that analyze the data and facilitates access to a collection of hosts with sufficient capacity to perform the analysis in situ.

Mobilyzer is a collaboration between Morley Mao’s group at the University of Michigan and David Choffnes’ group at Northeastern University. For more information, please visit the project page.

Funding

NSF

Related Publications

• Nikravesh, A., Yao, H., Xu, S., Choffnes, D., & Mao, Z. M. (2015, May). “Mobilyzer: An open platform for controllable mobile network measurements”. In Proceedings of the 13th Annual International Conference on Mobile Systems, Applications, and Services (pp. 389-404). DOI: 10.1145/2742647.2742670
• Nikravesh, A., Choffnes, D.R., Katz-Bassett, E., Mao, Z.M., Welsh, M. (2014). “Mobile Network Performance from User Devices: A Longitudinal, Multidimensional Analysis”. In Faloutsos, M., Kuzmanovic, A. (eds) Passive and Active Measurement. PAM 2014. Lecture Notes in Computer Science, vol 8362. Springer, Cham. DOI :10.1007/978-3-319-04918-2_2