1. Integrating base stations with a software- defined core network Bruno Hernández Zamora

2. Evolution of mobile traffic An unprecedented increase is expected in the demand for mobile data traffic that requires significant changes in the access network New technologies and networks (e.g. IoT) Reduction of cell size The elements of the access network need to become cheaper to install and maintain, easier to manage and update, and also more scalable 4/8/16 2

3. Enabling technologies Software-Defined Networking (SDN) Network architecture that is physically separated into forwarding and the control plane. Centralized control plane enables optimization opportunities. Network Functions Virtualization (NFV) Making use of IT virtualisation technology to standardize a wide variety of network components into a smaller subset of elements. 4/8/16 3

4. Evolved Packet Core 4/8/16 4

5. EPC – New Situation 4/8/16 5

6. Design tools GSMTAP header 4/8/16 6 Sub-types for LTE RRC defined in the GSMTAP header Virtual Machine A software program whose purpose is to provide the capability of executing programs as the physical machine would Linux network namespaces Each namespace can have its own routing table and run its own network processes

7. Simulation – VMs 4/8/16 7

8. Simulation – Packet bursts 4/8/16 8 Sequence of events in the asynchronous type of operation [1] [1] Boost.asio libraries overview (http://www.boost.org/doc/libs/1_55_0/doc/html/boost_asio/over view/core/basics.html) -Process time -Specify packet size (UDP payload 12B, total 48B) -Burst threshold 50ms Goal: measure processing limit of eNB

9. Results – Virtual Machine 4/8/16 9 λ (arrival rate) and µ (service rate) (s-1) in eNodeB, as a function of burst size (packets), in the downlink direction, from simulation run in VM Packet bursts sent downlink (from PDN host to eNodeB)

10. Results – Virtual Machine 4/8/16 10 λ (arrival rate) and µ (service rate) (s-1) in eNodeB, as a function of burst size (packets), in the uplink direction, from simulation run in VM Packet bursts sent uplink (from UE to eNodeB)

11. Results – No S/P GW 4/8/16 11 λ (arrival rate) (s-1) in eNodeB, as a function of burst size (packets), in the downlink direction, from simulation run in VM No interference from other applications µ (service rate) (s-1) in eNodeB, as a function of burst size (packets), in the downlink direction, from simulation run in VM

12. Results – No GTP 4/8/16 12 µ (service rate) (s-1) in eNodeB, as a function of burst size (packets), in the downlink direction, from simulation run in VM Effect of 16,67% in overhead reduction

13. Results – Hardware-enhanced VM 4/8/16 13 λ and µ (s-1) in eNodeB, as a function of burst size (packets), from simulation run in hardware-enhanced VM Increased stability for bigger packet bursts

14. Reference: Simulation – Namespaces 4/8/16 14