1) The document proposes LEADER, a collaborative edge- and SDN-assisted framework for HTTP adaptive video streaming. LEADER employs virtual network functions with transcoding capabilities at network edges to optimize video streaming quality of experience and network utilization. 2) An SDN controller runs an optimization model to determine the optimal location, action, and approach for fetching client-requested video qualities. A lightweight heuristic approach is also proposed. 3) An evaluation using a large-scale testbed of 250 clients, edge servers, and an SDN controller shows that LEADER improves average video bitrate, reduces quality switches and stalls, and increases perceived quality of experience over non-collaborative and default edge approaches. LE

1. LEADER : A Collaborative Edge- and SDN-Assisted
Framework for HTTP Adaptive Video Streaming
IEEE International Conference on Communications (ICC)
May 2022
reza.farahani@aau.at | https://athena.itec.aau.at/ | https://www.rezafarahani.me
Reza Farahani, Farzad Tashtarian, Christian Timmerer, Mohammad Ghanbari, and Hermann Hellwagner

2. Agenda
● Introduction
● Motivation
● Proposed solution
● Evaluation setup
● Experimental results
● Conclusion and Future work

3. Introduction
3

4. ● Video streaming traffic has become the primary type of traffic over the Internet.
○ It includes 53.72% of the total video traffic over the Internet [1]
● HTTP adaptive streaming (HAS) is one of the prominent technologies that delivers more than 51% of video
streams [1]
Introduction- Video Streaming
4
[1] Sandvine, “The Global Internet Phenamena Report,” White Paper, January 2022. [Online]. Available: https://www.sandvine.com/phenomena
https://bitmovin.com/dynamic-adaptive-streaming-http-mpeg-dash/

5. ✔ Pure client-based adaptation decision can lead clients to imperfect adaptations
◆ based on local parameters
◆ insufficient network information
Introduction- Network-Assisted Video Streaming
5
✔ Network-assisted Video Streaming for HAS:
◆ An in-network component with a broader view of the network is employed to assist video
players in making precise adaptation decisions
✔ Modern Networking paradigms, e.g.., SDN, NFV, edge computing have been
used for designing modern network-assisted frameworks

6. ✔ Conventional network architecture:
◆ Complex Network Devices
◆ Management Overhead
◆ Limited Scalability
✔ The control plane (forwarding decision) is decoupled from
the data plane (acts on the forwarding decision)
◆ Centralized Network Controller
◆ Standard communication Interface (OpenFlow),
◆ Programmable Open APIs
Introduction-Software-Defined Networking (SDN)
6
https://opennetworking.org/sdn-definition/

7. ✔ It is considered as a complementary technology to SDN
✔ NFV enables Virtual Network Functions (VNFs) to
◆ run over an open hardware platform
◆ Reduce OpEx, CapEx
◆ accelerate innovations
Introduction-Network Function Virtualization (NFV)
7
Router
Switch Load Balancer (LB)
Firewall
Virtualization Layer
VRouter VFirewall
VSwitch VLB
VNF VNF
VNF VNF

8. ✔ It provides storage and compute resources close to end-users at the network's edge, reducing
◆ network latency
◆ bandwidth consumption
✔ Edge servers include limited resources ( computation, storage, and bandwidth)
✔ Incorporating SDN, NFV, and Edge computing with HAS technology could
◆ optimize video streaming traffic and network utilization.
Introduction- Edge Computing
8

9. Motivation
9

10. 10
Motivation
✔ How to use edge resources efficiently to optimize users’ QoE and network utilization?
✔ How to design an edge- and SDN-assisted HAS framework for video optimization purposes?
✔ How to establish a collaboration between edge servers to use their potential idle resources
for serving HAS clients.
✔ How to design a network-assisted HAS scheme without client-side modification ?
✔ How we can implement and evaluate proposed approach in a large-scale testbed?
SDNN
F
V
HAS
E
d
g
e

11. Proposed Solution
11

12. The proposed framework
12
✔ This paper leverages the SDN, NFV, and edge computing technologies and proposes
◆ LEADER as a coLlaborative Edge- and SDN- Assisted framework for HTTP aDaptive vidEo
stReaming.
✔ LEADER employs VNFs with transcoding capability at the edge of an SDN-enabled network
✔ Edge servers are categorized into Local Edge Servers (LES) and Neighbor Edge Servers (NES)
Edge map
Comp Map
Requests
Cache map
Media
Media
Media

13. Central OptimizationModel
13
✔ SDN controller runs an MILP optimization model to respond to the following key questions:
1. Where is the optimal place (i.e., LES, NESs, CSs, or the origin server) in terms of the minimum serving
time for fetching each client’s requested content quality level from?
2. What is the optimal approach for answering to the requested quality level, i.e., fetch or transcode?
3. What is the optimal action to reach the requested quality?

14. Minimize Client serving time (i.e., fetching time plus transcoding time)
✔ Action Selection (AS) constraint
✔ Serving Time (ST) constraints
✔ CDN/Origin (CO) constraints
✔ Resource Consumption (RC)
Central OptimizationModel
14
✔ Constraints :
✔ Objective :

15. Distributed Heuristic Approach
15
✔ We propose a lightweight heuristic approach to:
1. remedy the high time complexity of the MILP model
2. Consider a bandwidth allocation strategy for shared links between each edge to other servers
✔ We propose the following time slot structure

16. SDN controller heuristic algorithm
16

17. Bandwidth Allocation Strategy
17

18. Edge Server Heuristic Algorithm
18

19. Evaluation setup
19

20. ✔ We design a large-scale cloud-based testbed, including 301 nodes (Xen virtual machines):
○ 250 clients
○ Four cache servers
○ 40 OpenFlow switches
○ 61 layer-2 links
○ An SDN controller (Floodlight)
○ Five edge servers (each edge server is responsible for 50 clients)
○ A video Dataset including:
■ Fifty video sequences (BBB with 150 segments)
■ 2 seconds segments
■ five representations (0.089, 0.262, 0.791, 2.4, 4.2 Mbps)
○ BOLA ABR algorithms
○ FFmpeg transcoder
○ Bandwidth monitoring (Floodlight Restful API)
○ LRU cache replacement policy
○ Zipf distribution video access popularity
Testbed
20

21. Experimental results
21

22. ✔ We evaluate the performance of LEADER compared to the following baseline systems
◆ Non Edge Collaborative (NECOL)
◆ Default Edge Collaborative (DECOL)
✔ The performance of the aforementioned approaches is evaluated through
◆ ASB: Average Segment Bitrate
◆ AQS: Average Number of Quality Switches
◆ ANS: Average Number of Stalls
◆ ASD: Average Stall Duration
◆ APQ: Average Perceived QoE calculated by ITU-T Rec.P.1203 mode 0
◆ CHR: Cache Hit Ratio
◆ ETR: Edge Transcoding Ratio
◆ BTL: Backhaul Traffic Load
Methods for Comparison and Metrics
22

23. QoE Results-- ASB and AQS
23

24. QoE Results-- ANS and ASD
24

25. QoE Results-- APQ
25

26. Network Utilization Results-- CHR and ETR
26

27. Network Utilization Results-- BTL
27

28. 28
Conclusion and Future work

29. ● This paper leverages the SDN, NFV and Edge computing paradigms to propose the
LEADER framework to provide high video QoE for HAS clients
● We design an edge collaborative architecture and formulate the problem as an
optimization model
● We propose a lightweight distributed heuristic approach including a bandwidth
allocation strategy
● We implement the proposed framework on a large-scale testbed consisting of 250
clients and conducts experiments for measuring QoE and Network Utilization metrics
● LEADER outperforms baseline schemes in terms of users’ QoE and the network
utilization by at least 22% and 13%, respectively
● Extending proposed Action tree, and employing learning approach are possible
future work directions.
Conclusion and Future Work
All rights reserved. ©2020 29

30. Thank you for your attention
reza.farahani@aau.at | https://www.rezafarahani.me | https://athena.itec.aau.at/
All rights reserved. ©2020
30