att_abstract={{Network function virtualization (NFV) promises significant cost savings, flexibility and ease of deployment. However, the lack of large scale operational deployments inhibits detailed understanding of the real world performance of virtualized network elements. For example, unlike data centers which have fairly straightforward switching, traditional telecom networks have a lot more complex interdependencies that can affect performance. In this paper, we use production grade,
software-based cellular network elements running on general purpose Linux servers to examine the combined effects of control and data planes on an LTE enhanced packet core (EPC). We focus our work on a ‘nodal-based’ architecture where each LTE core entity is operating in a separate virtual machine. Using a variety of realistic workloads derived from a cellular service provider, we quantify the impact of the nodal architecture on various metrics including physical processing, memory, IO, and
bandwidth resource requirements. Through experimentation, we discover that the performance bottlenecks are caused at the SGW as a result of the interdependencies between control and data plane entities. We vary factors like the number of subscribers, number of concurrent sessions, radio base stations, voice and data traffic activity to uncover several key scalability insights.}},
	att_authors={rj2124, je925g, vg7777},
	att_copyright_notice={{This version of the work is reprinted here with permission of IEEE for your personal use. Not for redistribution. The definitive version was published in 2015. {{, 2015-04-23}}
	att_tags={LTE, virtualization, EPC},
	author={Rittwik Jana and Jeffrey Erman and Vijay Gopalakrishnan and Ashok Sunder Rajan and Sameh Gobriel and Christian Maciocco and Kannan Babu Ramia and Sachin Kapur and Ajaypal Singh},
	institution={{2015 IEEE International Workshop on Local and Metropolitan Area Networks (LANMAN)}},
	title={{Understanding the bottlenecks in Virtualizing Cellular Core Network Functions}},