The introduction of iPhone by Apple Inc. in January 2007 has revolutionized the design of smart phones and ushered in a new era of smart phone “apps” and mobile computing. Now with wide adoption of smart phones and other mobile devices – e.g., iPad, Android-based and Blackberry-based “smart” pads (or tablets), cellular data traffic has grown tremendously in the past few years. This growth will be further precipitated by the increasing popularity of newer generations of smart-phones and mobile devices, along with many innovative “apps”(namely, various mobile applications and services, especially location-aware services) developed for these smart mobile devices. The rapid growth in cellular data traffic as well as the increasing demands for mobile data services have exerted enormous competitive pressure on cellular network service providers. In addition to quickly expanding the physical cellular infrastructure by adding capacity and increasing network coverage (e.g., through building new cell towers, etc.) and rapidly evolving their network architecture to meet user demands and expectations, these providers must also cope with many intricate and complex issues introduced by these new types of smart mobile devices and the “apps” and services running on them. How to effectively manage a large-scale cellular data network that is constantly evolving is therefore a daunting challenge that is not only important to cellular network service providers, but also to mobile data users and emerging mobile applications and services?
In this project we set out to identify, analyze and understand various challenging issues in, and develop mechanisms and tools for, effectively managing and trouble-shooting constantly evolving cellular data networks. The novelty of our research lies in several aspects. In close collaboration with our industrial research partners – in particular, AT&T Labs
, we are utilizing and leveraging vast amounts of heterogeneous sources of data – e.g., passive and active network measurement data and performance metrics, customer tickets, and so forth – obtained from operational cellular data networks. Guided by these real-world data, we are employing a data-driven, statistical machine learning
-based framework, and developing innovative machine learning and statistical inference techniques to analyze, mine and correlate various sources of data to uncover and identify various key factors (e.g., types of mobile devices, apps, usage patterns, mobility and locality, etc.) that contribute to network performance issues and affect user experiences and expectations. In particular, we are also taking advantage of customer tickets – namely, records of calls to customer support lines which provide first-hand indications of the issues and problems that users encounter – to build statistical models to make sense of customer tickets and understand the key factors and issues that affect user experiences and expectations. We are also utilizing customer reports and other user feedback to detect voice frauds, SMS spam and other suspicious activities and develop algorithms and methods for better securing the cellular networks. Building upon the analysis, models and understanding derived from our research, we will also develop effective mechanisms and tools for network management and trouble-shooting by correlating various heterogeneous data sources such as customer tickets, network measurement data and performance metrics. Finally, we will also develop innovative mobile management “apps” to provide on-the-spot, real-time user feedback to help diagnose and trouble-shoot network problems and improve user mobile experiences.
Undoubtedly, our research project constitutes only a first – albeit important step – toward gaining a better understanding of – as well as effectively managing and trouble-shooting – cellular data networks. There are numerous other challenges in operating and managing large-scale, evolving cellular data networks that will not and cannot be addressed in this project, due to its limited scope. Nonetheless, if successful, our research project will not only bring potential benefits to cellular service providers but also to hundreds of millions of users who are now increasingly dependent on mobile voice and data services. Our research can also provide valuable insights to inform the evolution and expansion of cellular networking infrastructure and services that are critical to the further development of emerging mobile and cloud computing.
Yu Jin, Nick Duffield, Alexandre Gerber, Patrick Haffner, Wen-Ling Hsu, Guy Jacobson, Subhabrata Sen, Shobha Venkataraman, and Zhi-Li Zhang, "Large-scale App-based Reporting of Customer Problems in Cellular Networks: Potential and Limitations", In Proc. ACM SIGCOMM Workshop on Measurements Up the STack (W-MUST) 2011, co-located with ACM SIGCOMM 2011, Toronto, ON, Canada, Aug 15 - Aug 19, 2011. [.pdf]
Zhenhua Li, Tieying Zhang, Yan Huang, Zhi-Li Zhang, Yafei Dai, "Cloud Transcoder: Bridging the Format and Resolution Gap between Internet Videos and Mobile Devices", In Proc. of Network and Operating System Support for Digital Audio and Video Workshop (NOSSDAV'12) Toronto, Ontario, CANADA, 2012. [.pdf]
Yu Jin, Nick Duffield, Alexandre Gerber, Patrick Haffner, Wen-Ling Hsu, Guy Jacobson, Subhabrata Sen, Shobha Venkataraman, Zhi-Li Zhang , "Characterizing Data Usage Patterns in a Large Cellular Network", In Proc. of ACM SIGCOMM CellNet Workshop, Helsinki, Finland, August 13, 2012. [.pdf]
Nan Jiang , Yu Jin, Ann Skudlark, Wen-Ling Hsu, Guy Jacobson, Siva Prakasam, Zhi-Li Zhang, "Isolating and Analyzing Fraud Activities in a Large Cellular Network via Voice Call Graph Analysis", In Proc. ACM Mobisys'12 Conference, Low Wood Bay, UK, 2012. [.pdf]
Keywords: cellular data networks, mobile data, network management, trouble-shooting, data-driven analysis and inference, data mining and machine learning.
The project is supported in part by