State: Seed funding via NSF I/UCRC, pending funding via NSF NeTS small grant
The proliferation of smart devices has placed considerable pressure on the data capacity of the cellular network. WiFi has emerged as the de facto technology for selectively augmenting capacity for the short term. Efforts in industry such as ANDSF (Access Network and Discovery Services Function) and Hotspot 2.0 (Passpoint) are poised to significantly improve the ability for users to automatically take advantage of nearby WiFi without direct intervention. Critically though, while WiFi generally tends to a positive for the user experience, user performance on WiFi is not always an improvement. Furthermore, automated roaming to WiFi has a strong potential to reflect poorly on the network carrier or an institutions. Roaming and dead spots in WiFi coverage coupled with the hard handoff necessary to roam back to cellular further compound challenges. Efforts such as Multi-Path TCP offer some promise but are still uncertain with regards to tuning and how / where data should be sent.
The primary challenge is how to effectively gauge the capacity and quality of the wireless link for a given mobile device at a given location. Traditional techniques for speed testing tend to be exceptionally heavyweight (ex. iperf-like mechanisms) moving potentially tens of megabits of traffic and taking on the order of seconds to complete. Lighter weight variants (PathChirp, Spruce, WBest) fare better with regards to cost but tend to suffer under packet aggregation optimizations introduced in more recent WiFi and cellular variants. For decisions such as roaming or more importantly, upstream decisions with regards to data routing, such information may only be of limited value by the time it is finally resolved.
Fast Mobile Network Characterization (FMNC) encompasses three sub-projects that include:
- FMNC Base: The focus of the base project for FMNC is to explore the extent to which one can reasonably characterize WiFi while requiring two orders magnitude less of data (100 KB or less), doing so in less than a quarter of a second, all without changes to the end client (beyond making a simple web get). We accomplish this task by leveraging our past work with RIPPS with an aggressive approach using sliced, structured, and reordered packets. We then create a sophisticated set of analytics at the server to extract the available bandwidth and other relevant characteristics. We have conducted some initial validations through the creation of an Android and iOS client as well as a complete server back-end.
- FMNC Cellular QoE: The same concepts from the WiFi variant of FMNC can potentially be leveraged for the purposes of assessing QoE when cellular links are involved. However, unlike WiFi which has reasonably consistent semantics for the creation of aggregated frames, the increased variation in construction of the RRB (Radio Resource Block) creates unique challenges as to assessing the available bandwidth of the cellular links. The focus of this sub-project is to explore the extent to which one can actively discern end-to-end performance (ex. streaming video) across cellular links utilizing similar concepts.
- FMNC Passive: The interesting observation afforded by FMNC base is that aggregated data frames often encapsulate a host of information with regards to queuing effects and link performance. The FMNC passive work seeks to explore if a purely passive technique could be constructed by merely observing existing traffic over brief periods. The ultimate goal would be to explore if even the limited WiFi scanning window would be useful enough to extract meaningful link performance.
|C||L. Song, A. Striegel, “Leveraging Frame Aggregation for Estimating WiFi Available Bandwidth,” to appear in Proc. of IEEE SECON, San Dieago, CA, June 2017.|
|C||L. Song, A. Striegel, “Leveraging Frame Aggregation to Improve Access Point Selection,” to appear in Proc. of INFOCOM International Workshop on Mobility Management in the Networks of the Future World (MobiWorld), Atlanta, GA, May 2017.|
[June 15, 2018] Great workshop in Amsterdam for NOSSDAV 2018. I had the pleasure of presenting my student Lixing’s paper entitled “SEWS: A Channel-Aware Stall-Free WiFi Video Streaming Mechanism” at the workshop in addition to serving as a session chair for the first session. Very well attended workshop and lots of great questions.
[April 24, 2018] Congrats to Lixing Song on getting his paper entitled “SEWS: A Channel-Aware Stall-Free WiFi Video Streaming Mechanism” accepted to appear at NOSSDAV 2018 in Amsterdam in June.
(March 14, 2017) Our paper entitled “Leveraging Frame Aggregation for Estimating WiFi Available Bandwidth” was accepted into IEEE SECON 2017 (26% acceptance rate). The work is the second one under the Fast Mobile Network Characterization umbrella. This work focuses on the potential to use reflected aggregation (client-mod free) as observed by A-MPDU frame characteristics for […]
(March 3, 2017) Our paper entitled “Leveraging Frame Aggregation to Improve Access Point Selection” has been accepted to appear at the INFOCOM International Workshop on Mobility Management in the Networks of the Future World (MobiWorld) workshop at INFOCOM 2017 to be held in Atlanta, GA in early May.
Presentations / Media
Software is available upon request to Prof. Striegel.
|Aaron Striegel||Lixing Song|