@techreport{TD:101273,
	att_abstract={{AT&T Labs Research
frastructures and through this, the experience for users using them.
Specifically, this paper presents our vision of how big data analytics can be used in real time to enhance the dependability of cellular network services. We demon- strate how big data can help in the design of adaptive techniques to reduce the incidence of voice or data dis- connections in cellular networks and mitigate their effect when they do occur. We show that such mitigation mech- anisms come with a cost that prohibits them from being turned on all the time. In order for them to be beneficial, real-time data analysis is necessary because network dis- connections depend not only on static factors (such as user locations that are prone to bad network connectiv- ity), but also on dynamic factors (such as current level of congestion in the cell, available radio resources, etc.). We analyze real data collected by a major cellular net- work and show that we can build a model that identifies conditions that are likely to lead to network disconnec- tions where real time mitigation mechanisms are benefi- cial.
2 Background
2.1 Network Architecture
Figure 1 shows our proposed architecture for the LTE network. The mobile device, called User Equipment (UE), is connected to a cell sector (which we will re- fer to as cell) in a base station, called eNodeB. A phys- ical base station can have multiple sectors, potentially covering different regions. Cellular traffic from the eN- odeB passes through Serving Gateway (S-Gateway) and Packet Data Network Gateway (PDN-Gateway) to exter- nal network (e.g., the Internet).
In order to identify (dynamic) conditions that can pre- dict voice or data drops with sufficient confidence, we add two components in the proposed architecture as shown in Figure 1: offline training and online predic-
There are many large infrastructures that instrument ev- erything from network performance metrics to user activ- ities. However, the collected data are generally used for long-term planning instead of improving reliability and user experience in real time. In this paper, we present our vision of how such collections of data can be used in real time to enhance the dependability of cellular net- work services. We first discuss mitigation mechanisms that can be used to improve reliability, but incur a high cost which prohibit them to be used except in certain con- ditions. We present two case studies where analyses of real cellular network traffic data show that we can iden- tify these conditions.}},
	att_authors={kj2681, rp267p},
	att_categories={},
	att_copyright={{ACM}},
	att_copyright_notice={{(c) ACM, 2013. This is the author's version of the work. It is posted here by permission of ACM for your personal use. Not for redistribution. The definitive version was published in 2013] {{, 2013-11-03}}.
}},
	att_donotupload={},
	att_private={false},
	att_projects={},
	att_tags={},
	att_techdoc={true},
	att_techdoc_key={TD:101273},
	att_url={http://web1.research.att.com:81/techdocs_downloads/TD:101273_DS1_2013-08-15T18:28:08.612Z.pdf},
	author={Kaustubh Joshi and Rajesh Panta and Nawanol Theera-Ampornpunt and Saurabh Bagchi},
	institution={{9th ACM Workshop on Hot Topics in Dependable Systems}},
	month={November},
	title={{Using Big Data for More Dependability: A Cellular Network Tale}},
	year=2013,
}