Mobile Edge Computing in Cellular Network

1. Mobile Edge Computing in Cellular Network Antony Franklin Networked Wireless Systems Lab IIT Hyderabad

2. Evolution of Mobile Edge Computing • Cloud computing [2000] • Cloudlets [CMU 2009] • Fog computing [CISCO 2012] • Mobile Edge Computing [ETSI 2014] • Multi-access Edge Computing

3. Mobile Edge Computing • Edge computing is one of the key enabler of cloud computing by keeping services at edge or at first hop • MEC offers IT service and cloud-computing capabilities at the edge of the mobile network in an environment that is characterized by proximity, ultra-low latency and high bandwidth. Furthermore, it provides exposure to real-time radio network and context information – ETSI (European Telecommunications Standards Institute) • The primary goal of edge computing is to reduce network congestion and improve application performance by executing related task processing closer to the end user, improving the delivery of content and applications to those users

4. Characteristics of MEC • Proximity [1 or 2 hops] • Location awareness • High throughput [1-10 Gbps] • Low latency [1 millisecond] • High reliability [99.9999% availability] • Energy efficiency [90% reduction in energy usage] • Less backhaul network congestion

5. MEC Use Cases

6. MEC based AR/VR System • Can choose rendering pipeline either in a ME application or on the UE • Can choose to offload part of computation • Relocation of application

7. Video Caching/Acceleration through MEC • Content is consumed at about the same time in the same geographical area (Store the popular content locally) • Local content caching, saves the backhaul requirement • Quick download of the content improved QoE of video

8. MEC based V2X Infrastructure • Roadside unit is intended to increase the safety, efficiency, and convenience • Data from vehicles and sensors to recognize high-risk situations • Tight latency requirements • Application can be deployed on ME hosts to provide roadside functionality

9. MEC based IoT Gateway • IoT Gateway application deployed at MEC server • IoT vertical specific data analytics at the edge • Data aggregation at the edge

10. Middlebox based MEC deployment in LTE • Bump-in the wire approach • No modifications required on the core network and base station

11. Middlebox based MEC deployment in LTE • Intercept and forward the GTP packet between eNB and S-GW • MEC application servers serves the packet embedded in GTP • Traffic Redirection via Proxy ARP • Stateful tracking of GPT tunnel

12. SDN based MEC deployment in LTE • LLMEC developed by Eurecom for enabling Low Latency Edge Application • Use of SDN to implement Control and User Plane Split (CUPS) • Use of northbound APIs for traffic redirection • Moving PGW functionalities at OpenVSwitch for traffic steering

13. SDN based MEC deployment in LTE

14. Consolidated Caching and Cache Splitting • Consolidated Caching: No replication, only one copy of video in the cache network. More videos are stored in the network but increase in delay • Cache Splitting: Logical splitting of cache to store complete and initial segments of the video. Helps in reducing the delay

15. Consolidated Caching and Cache Splitting

16. MEC in 5G • 5G provides higher data rate than 4G (1000x bandwidth per unit area)  more back haul traffic in the 5G core • 5G RAN provides low RAN latency (1 ms)  Backhaul is the bottleneck for the low latency services

17. MEC in 5G • UPFs are distributed and configurable data plane from the MEC system perspective

18. MEC support functions in 5G • UPF (Re)selection - The 5G Core Network (re)selects UPF to route the user traffic to the Local Area Data Network (LADN) • Traffic influence by application function - The AF can influence UPF (re)selection either communicating with the PCF or NEF • Local routing and traffic steering - Traffic steering through Uplink Classifiers at UPF that operate on a set of traffic filters matching the steered traffic • Session and service continuity - Different SSC modes [IP address modifications] are specified to enable UE and application mobility • Network Capability exposure - UE’s information like IP address, location, radio quality will be exposed by 5G core network • QoS and Charging using PCF policies - QoS Control and Charging policies for the traffic routed through LADN while UPF tracks data usage

19. MEC Deployment Scenarios in 5G 5G Control Plane NEF AUSF UDM NRF PCF AMF SMF N4 N2 Local UPF support with local area data network (LADN) UE RAN Local UPF LADN N3 N6

20. MEC Deployment Scenarios in 5G 5G Control Plane NEF AUSF UDM NRF PCF AMF SMF N4 N2 N4 N6 N3 N9 UE RAN UPF PSA UPF1 DN Multiple UPF support: One UPF for MEC and one UPF for DN N9 N6 Local UPF MEC

21. MEC Deployment Scenarios in 5G 5G Control Plane NEF AUSF UDM NRF PCF AMF SMF N4 N2 Single UPF Connects to Both MEC and DN UE RAN UPF MEC N3 N6 IP N6 DN

22. MEC Development at IIT Hyderabad • Any third party application can be deployed on the 5G network • Both Trusted and Untrusted MEC platform integration with 5G • Both Single and Multiple UPF scenarios

23. Conclusion • Mobile Edge Computing is the way to meet the requirement of 5G applications • Middle-box and SDN are the way to implement MEC in 4G • 5G is designed with the notion of MEC in the mobile networks. Lot of support functionalities defined in the standard • Many 5G applications that require high bandwidth and/or low latency requirement can be realized with MEC

24. Thank you !!