home
201302_Towerscan-front

When Cell Towers Fail: Quantifying the Customer Impact

by: He Yan, Zihui Ge, Matt Osinski, Jennifer Yates, March 6, 2013

When cell towers fail, what happens to customers? In a resilient network where customers may simply move from a failed tower to another one nearby, the answer is not always clear. And it can’t be captured by manual assessments that rely only on counting the number of failed towers and estimating population density, and thus ignore how an outage may be dispersed over a wide area.

What’s needed is an in-depth analysis of network data. It’s why AT&T Researchers created a tower-outage analyzer, which looks at the complex interactions among customers and cell towers to intelligently assess the customer experience across an entire impact area. Read more.

Two Bowls, Some Pebbles Signal Digital Availability

201307-yarosh-bowls

Imagine arriving home and signaling your digital availability by moving pebbles from one bowl to another: one bowl for close friends only (or for text messages only), another for wider availability. It's a refreshingly simple use of physical objects to maintain control in a plugged-in age. And it is the idea behind Lana Yarosh's Availabowls project, which works by attaching programmable RFID tags to pebbles. For more information about Availabowls, see this article.


Paper on IP Traffic Matrices Stands Test of Time

Test-of-time_213x108

For the second year in a row, the ACM SIGMETRICS conference conferred its Test of Time Award on a paper co-authored by Nick Duffield. This year the award—given to a previous conference paper whose impact is still felt 10 years later—went to the 2003 Fast Accurate Computation of Large-Scale IP Traffic Matrices from Link Load for its "novel, remarkably fast, and accurate method for practical and rapid inference of traffic matrices in IP networks from link load measurements."


Mazin Gilbert Interviewed in Popular Mechanics

201307_interview_mazin

Machine speech translation is hard to do; it requires recognizing the language, transcribing it, doing the translation—all while the person is talking. Processing requirements are enormous.

In this Popular Mechanics interview, Mazin Gilbert describes how AT&T Research is incorporating machine learning and cloud technology to enhance its speech technologies—representing over 30 years of research—and make smooth, seamless translation a reality.