SlideShare a Scribd company logo
1 of 28
Download to read offline
All rights reserved. ©2020
1
Quality Optimization of Live Streaming Services
over HTTP with Reinforcement Learning
The IEEE Global Communications Conference (GLOBECOM)
7 -11 December 2021, Madrid, Spain
Farzad Tashtarian, Reyhane Falanji, Abdelhak Bentaleb, Alireza Erfanian, Peyman S. Mashhadi,
Christian Timmerer, Hermann Hellwagner, and Roger Zimmermann
Christian Doppler laboratory ATHENA | Klagenfurt University | Austria
farzad.tashtarian@aau.at | https://athena.itec.aau.at/
All rights reserved. ©2020
● Introduction
● Proposed approach
● Results
● Conclusion and future work
● Q&A
Agenda
All rights reserved. ©2020
2
All rights reserved. ©2020
Introduction
● Recent years have seen tremendous growth in HTTP adaptive live video
traffic over the Internet.
● HTTP Adaptive Streaming (HAS) has become the de-facto solution
allowing Over-The-Top(OTT) services to deliver an acceptable Quality of
Experience (QoE)
● Dynamic network conditions and diverse request patterns
○ A large overhead and significant increase in time complexity
3
4
HAS Player
Video Distribution Network
Internet
Video Contribution
CDN Server
CDN Server
CDN Server
Origin Server
ABR
Encoder
Live
Source
HAS Player
HAS Player
5
HAS Player
Video Distribution Network
Internet
Video Contribution
CDN Server
CDN Server
CDN Server
Origin Server
ABR
Encoder
Live
Source
HAS Player
HAS Player
6
HAS Player
Video Distribution Network
Internet
Video Contribution
CDN Server
CDN Server
CDN Server
Origin Server
ABR
Encoder
Live
Source
HAS Player
HAS Player
HTTP Request
for blue segment
HTTP Request
for red segment
7
HAS Player
Video Distribution Network
Internet
Video Contribution
CDN Server
CDN Server
CDN Server
Origin Server
ABR
Encoder
Live
Source
HAS Player
HAS Player
HTTP Response
for blue segment
HTTP Response
for red segment
8
Video Distribution Network
Internet
Video Contribution
CDN Server
Origin Server
ABR
Encoder
Live
Source
HAS Player
HTTP Response
for blue segment
HTTP Response
for red segment
How to increase clients’ QoE by considering :
1- Network Bandwidth between users and CDN server
2- Number of requests for a similar channel and quality
3- Different serving methods:
- Fetch from CDN/origin server
- Transcoding from higher quality to lower one
- Serving with lower quality
All rights reserved. ©2020
Problem Definition
How do we process and serve numerous clients’
requests, demanding different live channels with
various bitrates, at the network edge to elevate the
clients’ perceived QoE in a network with adverse
bandwidth conditions?
9
10
Video Distribution Network
Internet
Video Contribution
CDN Server
Origin Server
ABR
Encoder
Live
Source
HAS Player
Solution: ROPL,
● A learning-based client request management
solution at the edge
● leverage the deep reinforcement learning,
● serve requests of concurrent users joining
various HTTP-based live video channels
RL-based Virtual
Reverse Proxy
(RVP)
HTTP Request
for blue segment
11
Video Distribution Network
Internet
Video Contribution
CDN Server
Origin Server
ABR
Encoder
Live
Source
HAS Player
Solution: ROPL,
● A learning-based client request management
solution at the edge
● leverage the deep reinforcement learning,
● serve requests of concurrent users joining
various HTTP-based live video channels
RL-based Virtual
Reverse Proxy
(RVP)
HTTP Response
for blue segment
All rights reserved. ©2020
ROPL Architecture
● The RVP consists of the following modules:
○ service manager(SM),
○ bitrate classifier (BC),
○ bandwidth monitor (BM),
○ partial cache,
○ deep reinforcement learning (DRL)agent.
12
All rights reserved. ©2020
Deep Reinforcement Learning (DRL) Agent
● Proximal Policy Optimization (PPO)
● Policy gradient-based algorithm with an actor-critic approach
● Markov Decision Process (MDP) model
● The translation of our scenario into an MDP model is as
follows:
○ State Space
○ Action Space
○ Policy
○ Reward Function
13
All rights reserved. ©2020
All rights reserved. ©2020
DRL Agent- State Space
14
Observation Space
indicates the number of the same requests for bitrate j
(considering an ascending order) of the live channel i
is the available bandwidth given from the SM module
All rights reserved. ©2020
DRL Agent- Action Space
15
is a set of actions that can be performed by the DRL agent
A set of selected actions by the DRL agent at time step τ
Indicates the selected action for aggregated requests in the jth
queue of the ith live channel with number
16
ACT1: fetching the requested segment s with bitrate j directly
from the remote CDN/origin server;
ACT2: serving segments with bitrate j* demanded by request i*
in the same time step, where j* < j and action of i* is ACT#1
ACT3: serving by transcoding segment s from a higher bitrate
j* demanded by request i* where ACT#1 is selected for i*
ACT4: do nothing
DRL Agent- Action Space ...
All rights reserved. ©2020
DRL Agent- Policy
● Policy refers to the logic based on which the agent takes
action according to a particular observation.
● The policy is optimized by learning from the environment and
updates based on reward signals.
● A desired convergence of policy happens when the agent learns
to take the most rewarding action with the highest conditional
probability, and a policy update does not change the current
one.
17
All rights reserved. ©2020
DRL Agent- Reward Function
18
❏ An undeniable role in the DRL
❏ Guides the agent towards the following purpose:
serving requests with the lowest cost and the maximum QoE
The proposed reward function consists of
● Serving cost
● Violation penalty
DRL Agent- Reward Function - Serving Cost
19
Serving cost C1-C4 for applying ACK1-ACK4 respectively.
by considering the cost for serving
a requested bitrate with a lower
one, coefficient α2 is selected
the required
transcoding time
A coefficient regarding
the cost of bandwidth for
applying ACT#1
A coefficient regarding the cost of
computational resources for applying
ACT#3,
Serving by fetch Serving by lower bitrates
Serving by transcoding
All rights reserved. ©2020
DRL Agent- Reward Function - Violation Penalty
20
For every action set in a timestep, penalty signals are encountered as
many as the number of violations according to the following constraints:
● CON#1: empty request
○ whenever any action other than ACT#4 is taken for an empty request
● CON#2: bandwidth
○ when a fetch action (ACT#1) is taken, but the bandwidth limit is
violated.
● CON#3: existing of lower bitrate
○ when the agent decides to serve requests in a queue by a lower
bitrate, however ACT#1 has not been taken for a queue that
contains request(s) with similar segment number but lower bitrate;
● CON#4: existing of higher bitrate
○ when a queue is supposed to be served by transcoding from a
higher bitrate version of the same segment, but no queue with a
higher bitrate request has performed a successful fetch action
(ACT#1);
● CON#5: non-empty queue
○ a “do nothing” decision for a non-empty queue.
21
DRL Agent- Reward Function
the normalized
total cost in time
step τ
the normalized
total penalty in time
step τ
sum of violation in
time step τ
sum of actions’ cost and
violation in time step τ
22
Heuristic Post-Processing Algorithm
Prevent applying invalid actions on requests
23
Performance Evaluation
We conduct the performance evaluation in two modes:
● Simulation scenarios
○ six scenarios for different numbers of live channels, players, and various
amounts of bandwidth
● Real-world scenarios
○ RVP in Python to serve requests received from the goDASH player
○ HTTP origin server on an AWS virtual machine located in the Frankfurt
zone
○ CH1(Big Buck Bunny), CH2(Tears of Steel), CH3(Sintel)
CH1: {180p@0.25Mbps, 360p@1.1Mbps, 414p@1.82Mbps, 720p@3.0Mbps, 1080p@3.89Mbps}
CH2: {216p@0.39Mbps, 360p@1.1Mbps, 414p@1.82Mbps, 720p@2.38Mbps, 1080p@3.89Mbps}
CH3: {288p@0.57Mbps, 414p@1.82Mbps, 720p@2.38Mbps, 1080p@3.89Mbps, 1080p@4.3Mbps}
24
Simulation Experiments
Frequency of penalties (a) and taken actions (b) for different
scenarios using ROPL model in simulation
Simulation Experiments ...
25
Resulted reward (a) and loss (b) using ROPL model in simulations
Real-world Experiment
26
Achieved average QoE for each channel clients and different total
bandwidth configurations using ROPL in real-world experiments.
All rights reserved. ©2020
Conclusion and Future Work
● We proposed ROPL,
○ a reinforcement learning approach
○ optimizing HTTP-based live streaming
○ an edge based mechanism
○ trace-driven simulations and a real-world setup
○ outperforms existing heuristic-based approaches in terms of
QoE, with a factor up to 3.7×
● Extend ROPL to support multiple RVPs
27
All rights reserved. ©2020
2
Quality Optimization of Live Streaming Services
over HTTP with Reinforcement Learning
The IEEE Global Communications Conference (GLOBECOM)
7 -11 December 2021, Madrid, Spain
Farzad Tashtarian, Reyhane Falanji, Abdelhak Bentaleb, Alireza Erfanian, Peyman S. Mashhadi,
Christian Timmerer, Hermann Hellwagner, and Roger Zimmermann
Christian Doppler laboratory ATHENA | Klagenfurt University | Austria
farzad.tashtarian@aau.at | https://athena.itec.aau.at/

More Related Content

What's hot

ES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video Streaming
ES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video StreamingES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video Streaming
ES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video Streaming
Alpen-Adria-Universität
 
CSDN: CDN-Aware QoE Optimization in SDN-Assisted HTTP Adaptive Video Streaming
CSDN: CDN-Aware QoE Optimization in SDN-Assisted HTTP Adaptive Video StreamingCSDN: CDN-Aware QoE Optimization in SDN-Assisted HTTP Adaptive Video Streaming
CSDN: CDN-Aware QoE Optimization in SDN-Assisted HTTP Adaptive Video Streaming
Alpen-Adria-Universität
 
CAdViSE or how to find the Sweet Spots of ABR Systems
CAdViSE or how to find the Sweet Spots of ABR SystemsCAdViSE or how to find the Sweet Spots of ABR Systems
CAdViSE or how to find the Sweet Spots of ABR Systems
Alpen-Adria-Universität
 
On Optimizing Resource Utilization in AVC-based Real-time Video Streaming
On Optimizing Resource Utilization in AVC-based Real-time Video StreamingOn Optimizing Resource Utilization in AVC-based Real-time Video Streaming
On Optimizing Resource Utilization in AVC-based Real-time Video Streaming
Alpen-Adria-Universität
 
CAdViSE: Cloud based Adaptive Video Streaming Evaluation Framework for the Au...
CAdViSE: Cloud based Adaptive Video Streaming Evaluation Framework for the Au...CAdViSE: Cloud based Adaptive Video Streaming Evaluation Framework for the Au...
CAdViSE: Cloud based Adaptive Video Streaming Evaluation Framework for the Au...
Alpen-Adria-Universität
 
Quality impact of scalable video coding tunneling for media aware content del...
Quality impact of scalable video coding tunneling for media aware content del...Quality impact of scalable video coding tunneling for media aware content del...
Quality impact of scalable video coding tunneling for media aware content del...
Alpen-Adria-Universität
 

What's hot (20)

HTTP Adaptive Streaming – Where Is It Heading?
HTTP Adaptive Streaming – Where Is It Heading?HTTP Adaptive Streaming – Where Is It Heading?
HTTP Adaptive Streaming – Where Is It Heading?
 
ES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video Streaming
ES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video StreamingES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video Streaming
ES-HAS: An Edge- and SDN-Assisted Framework for HTTP Adaptive Video Streaming
 
Docker-Based Evaluation Framework for Video Streaming QoE in Broadband Networks
 Docker-Based Evaluation Framework for Video Streaming QoE in Broadband Networks Docker-Based Evaluation Framework for Video Streaming QoE in Broadband Networks
Docker-Based Evaluation Framework for Video Streaming QoE in Broadband Networks
 
Scalable High Efficiency Video Coding based HTTP Adaptive Streaming over QUIC...
Scalable High Efficiency Video Coding based HTTP Adaptive Streaming over QUIC...Scalable High Efficiency Video Coding based HTTP Adaptive Streaming over QUIC...
Scalable High Efficiency Video Coding based HTTP Adaptive Streaming over QUIC...
 
Bandwidth Prediction in Low-Latency Chunked Streaming
Bandwidth Prediction in Low-Latency Chunked StreamingBandwidth Prediction in Low-Latency Chunked Streaming
Bandwidth Prediction in Low-Latency Chunked Streaming
 
A Distributed Delivery Architecture for User Generated Content Live Streaming...
A Distributed Delivery Architecture for User Generated Content Live Streaming...A Distributed Delivery Architecture for User Generated Content Live Streaming...
A Distributed Delivery Architecture for User Generated Content Live Streaming...
 
20 Years of Streaming in 20 Minutes
20 Years of Streaming in 20 Minutes20 Years of Streaming in 20 Minutes
20 Years of Streaming in 20 Minutes
 
CSDN: CDN-Aware QoE Optimization in SDN-Assisted HTTP Adaptive Video Streaming
CSDN: CDN-Aware QoE Optimization in SDN-Assisted HTTP Adaptive Video StreamingCSDN: CDN-Aware QoE Optimization in SDN-Assisted HTTP Adaptive Video Streaming
CSDN: CDN-Aware QoE Optimization in SDN-Assisted HTTP Adaptive Video Streaming
 
LwTE: Light-weight Transcoding at the Edge
LwTE: Light-weight Transcoding at the EdgeLwTE: Light-weight Transcoding at the Edge
LwTE: Light-weight Transcoding at the Edge
 
EADAS: Edge Assisted Adaptation Scheme for HTTP Adaptive Streaming
EADAS: Edge Assisted Adaptation Scheme for HTTP Adaptive StreamingEADAS: Edge Assisted Adaptation Scheme for HTTP Adaptive Streaming
EADAS: Edge Assisted Adaptation Scheme for HTTP Adaptive Streaming
 
INCEPT: Intra CU Depth Prediction for HEVC
INCEPT: Intra CU Depth Prediction for HEVCINCEPT: Intra CU Depth Prediction for HEVC
INCEPT: Intra CU Depth Prediction for HEVC
 
CAdViSE or how to find the Sweet Spots of ABR Systems
CAdViSE or how to find the Sweet Spots of ABR SystemsCAdViSE or how to find the Sweet Spots of ABR Systems
CAdViSE or how to find the Sweet Spots of ABR Systems
 
On Optimizing Resource Utilization in AVC-based Real-time Video Streaming
On Optimizing Resource Utilization in AVC-based Real-time Video StreamingOn Optimizing Resource Utilization in AVC-based Real-time Video Streaming
On Optimizing Resource Utilization in AVC-based Real-time Video Streaming
 
Video Coding for Large-Scale HTTP Adaptive Streaming Deployments: State of th...
Video Coding for Large-Scale HTTP Adaptive Streaming Deployments: State of th...Video Coding for Large-Scale HTTP Adaptive Streaming Deployments: State of th...
Video Coding for Large-Scale HTTP Adaptive Streaming Deployments: State of th...
 
Understanding Quality of Experience of Heuristic-based HTTP Adaptive Bitrate ...
Understanding Quality of Experience of Heuristic-based HTTP Adaptive Bitrate ...Understanding Quality of Experience of Heuristic-based HTTP Adaptive Bitrate ...
Understanding Quality of Experience of Heuristic-based HTTP Adaptive Bitrate ...
 
FaME-ML: Fast Multirate Encoding for HTTP Adaptive Streaming Using Machine Le...
FaME-ML: Fast Multirate Encoding for HTTP Adaptive Streaming Using Machine Le...FaME-ML: Fast Multirate Encoding for HTTP Adaptive Streaming Using Machine Le...
FaME-ML: Fast Multirate Encoding for HTTP Adaptive Streaming Using Machine Le...
 
SLFC: Scalable Light Field Coding
SLFC: Scalable Light Field CodingSLFC: Scalable Light Field Coding
SLFC: Scalable Light Field Coding
 
Press Release of 131st WG11 (MPEG) Meeting
Press Release of 131st WG11 (MPEG) MeetingPress Release of 131st WG11 (MPEG) Meeting
Press Release of 131st WG11 (MPEG) Meeting
 
CAdViSE: Cloud based Adaptive Video Streaming Evaluation Framework for the Au...
CAdViSE: Cloud based Adaptive Video Streaming Evaluation Framework for the Au...CAdViSE: Cloud based Adaptive Video Streaming Evaluation Framework for the Au...
CAdViSE: Cloud based Adaptive Video Streaming Evaluation Framework for the Au...
 
Quality impact of scalable video coding tunneling for media aware content del...
Quality impact of scalable video coding tunneling for media aware content del...Quality impact of scalable video coding tunneling for media aware content del...
Quality impact of scalable video coding tunneling for media aware content del...
 

Similar to Quality Optimization of Live Streaming Services over HTTP with Reinforcement Learning

LwTE-Live: Light-weight Transcoding at the Edge for Live Streaming
LwTE-Live: Light-weight Transcoding at the Edge for Live StreamingLwTE-Live: Light-weight Transcoding at the Edge for Live Streaming
LwTE-Live: Light-weight Transcoding at the Edge for Live Streaming
Alpen-Adria-Universität
 
PLNOG 13: Bart Salaets: Optimising TCP in today’s changing network environment
PLNOG 13: Bart Salaets: Optimising TCP in today’s changing network environmentPLNOG 13: Bart Salaets: Optimising TCP in today’s changing network environment
PLNOG 13: Bart Salaets: Optimising TCP in today’s changing network environment
PROIDEA
 
How to Optimize Dynamic Adaptive Video Streaming? Challenges and Solutions
How to Optimize Dynamic Adaptive Video Streaming? Challenges and SolutionsHow to Optimize Dynamic Adaptive Video Streaming? Challenges and Solutions
How to Optimize Dynamic Adaptive Video Streaming? Challenges and Solutions
Alpen-Adria-Universität
 
CSDN_ CDN-Aware QoE Optimization inSDN-Assisted HTTP Adaptive Video Streaming...
CSDN_ CDN-Aware QoE Optimization inSDN-Assisted HTTP Adaptive Video Streaming...CSDN_ CDN-Aware QoE Optimization inSDN-Assisted HTTP Adaptive Video Streaming...
CSDN_ CDN-Aware QoE Optimization inSDN-Assisted HTTP Adaptive Video Streaming...
Reza Farahani
 
Policy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Policy-Driven Dynamic HTTP Adaptive Streaming Player EnvironmentPolicy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Policy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Alpen-Adria-Universität
 
Policy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Policy-Driven Dynamic HTTP Adaptive Streaming Player EnvironmentPolicy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Policy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Minh Nguyen
 

Similar to Quality Optimization of Live Streaming Services over HTTP with Reinforcement Learning (20)

PLNOG 7: Emil Gągała, Sławomir Janukowicz - carrier grade NAT
PLNOG 7: Emil Gągała,  Sławomir Janukowicz - carrier grade NAT PLNOG 7: Emil Gągała,  Sławomir Janukowicz - carrier grade NAT
PLNOG 7: Emil Gągała, Sławomir Janukowicz - carrier grade NAT
 
Optimizing QoE and Latency of Live Video Streaming Using Edge Computing a...
Optimizing  QoE and Latency of  Live Video Streaming Using  Edge Computing  a...Optimizing  QoE and Latency of  Live Video Streaming Using  Edge Computing  a...
Optimizing QoE and Latency of Live Video Streaming Using Edge Computing a...
 
LwTE-Live: Light-weight Transcoding at the Edge for Live Streaming
LwTE-Live: Light-weight Transcoding at the Edge for Live StreamingLwTE-Live: Light-weight Transcoding at the Edge for Live Streaming
LwTE-Live: Light-weight Transcoding at the Edge for Live Streaming
 
Microservices summit talk 1/31
Microservices summit talk   1/31Microservices summit talk   1/31
Microservices summit talk 1/31
 
PLNOG 13: Bart Salaets: Optimising TCP in today’s changing network environment
PLNOG 13: Bart Salaets: Optimising TCP in today’s changing network environmentPLNOG 13: Bart Salaets: Optimising TCP in today’s changing network environment
PLNOG 13: Bart Salaets: Optimising TCP in today’s changing network environment
 
Bringing Learnings from Googley Microservices with gRPC - Varun Talwar, Google
Bringing Learnings from Googley Microservices with gRPC - Varun Talwar, GoogleBringing Learnings from Googley Microservices with gRPC - Varun Talwar, Google
Bringing Learnings from Googley Microservices with gRPC - Varun Talwar, Google
 
IEEEGlobecom'22-OL-RICHTER.pdf
IEEEGlobecom'22-OL-RICHTER.pdfIEEEGlobecom'22-OL-RICHTER.pdf
IEEEGlobecom'22-OL-RICHTER.pdf
 
Chapter04
Chapter04Chapter04
Chapter04
 
Charles sonigo - Demuxed 2018 - How to be data-driven when you aren't Netflix...
Charles sonigo - Demuxed 2018 - How to be data-driven when you aren't Netflix...Charles sonigo - Demuxed 2018 - How to be data-driven when you aren't Netflix...
Charles sonigo - Demuxed 2018 - How to be data-driven when you aren't Netflix...
 
(Slides) P2P video broadcast based on per-peer transcoding and its evaluatio...
(Slides) P2P video broadcast based on per-peer transcoding and its evaluatio...(Slides) P2P video broadcast based on per-peer transcoding and its evaluatio...
(Slides) P2P video broadcast based on per-peer transcoding and its evaluatio...
 
How to Optimize Dynamic Adaptive Video Streaming? Challenges and Solutions
How to Optimize Dynamic Adaptive Video Streaming? Challenges and SolutionsHow to Optimize Dynamic Adaptive Video Streaming? Challenges and Solutions
How to Optimize Dynamic Adaptive Video Streaming? Challenges and Solutions
 
CCNA (R & S) Module 01 - Introduction to Networks - Chapter 9
CCNA (R & S) Module 01 - Introduction to Networks - Chapter 9CCNA (R & S) Module 01 - Introduction to Networks - Chapter 9
CCNA (R & S) Module 01 - Introduction to Networks - Chapter 9
 
Internet Video
Internet VideoInternet Video
Internet Video
 
ACM NOSSDAV'21-ES-HAS_ An Edge- and SDN-Assisted Framework for HTTP Adaptive ...
ACM NOSSDAV'21-ES-HAS_ An Edge- and SDN-Assisted Framework for HTTP Adaptive ...ACM NOSSDAV'21-ES-HAS_ An Edge- and SDN-Assisted Framework for HTTP Adaptive ...
ACM NOSSDAV'21-ES-HAS_ An Edge- and SDN-Assisted Framework for HTTP Adaptive ...
 
CSDN_ CDN-Aware QoE Optimization inSDN-Assisted HTTP Adaptive Video Streaming...
CSDN_ CDN-Aware QoE Optimization inSDN-Assisted HTTP Adaptive Video Streaming...CSDN_ CDN-Aware QoE Optimization inSDN-Assisted HTTP Adaptive Video Streaming...
CSDN_ CDN-Aware QoE Optimization inSDN-Assisted HTTP Adaptive Video Streaming...
 
Policy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Policy-Driven Dynamic HTTP Adaptive Streaming Player EnvironmentPolicy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Policy-Driven Dynamic HTTP Adaptive Streaming Player Environment
 
DHCP
DHCPDHCP
DHCP
 
2017 01-31-cgns
2017 01-31-cgns2017 01-31-cgns
2017 01-31-cgns
 
Reconsider TCPdump for Modern Troubleshooting
Reconsider TCPdump for Modern TroubleshootingReconsider TCPdump for Modern Troubleshooting
Reconsider TCPdump for Modern Troubleshooting
 
Policy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Policy-Driven Dynamic HTTP Adaptive Streaming Player EnvironmentPolicy-Driven Dynamic HTTP Adaptive Streaming Player Environment
Policy-Driven Dynamic HTTP Adaptive Streaming Player Environment
 

More from Alpen-Adria-Universität

Optimal Quality and Efficiency in Adaptive Live Streaming with JND-Aware Low ...
Optimal Quality and Efficiency in Adaptive Live Streaming with JND-Aware Low ...Optimal Quality and Efficiency in Adaptive Live Streaming with JND-Aware Low ...
Optimal Quality and Efficiency in Adaptive Live Streaming with JND-Aware Low ...
Alpen-Adria-Universität
 
Content-adaptive Video Coding for HTTP Adaptive Streaming
Content-adaptive Video Coding for HTTP Adaptive StreamingContent-adaptive Video Coding for HTTP Adaptive Streaming
Content-adaptive Video Coding for HTTP Adaptive Streaming
Alpen-Adria-Universität
 
Evaluation of Quality of Experience of ABR Schemes in Gaming Stream
Evaluation of Quality of Experience of ABR Schemes in Gaming StreamEvaluation of Quality of Experience of ABR Schemes in Gaming Stream
Evaluation of Quality of Experience of ABR Schemes in Gaming Stream
Alpen-Adria-Universität
 
Energy Consumption in Video Streaming: Components, Measurements, and Strategies
Energy Consumption in Video Streaming: Components, Measurements, and StrategiesEnergy Consumption in Video Streaming: Components, Measurements, and Strategies
Energy Consumption in Video Streaming: Components, Measurements, and Strategies
Alpen-Adria-Universität
 
Exploring the Energy Consumption of Video Streaming: Components, Challenges, ...
Exploring the Energy Consumption of Video Streaming: Components, Challenges, ...Exploring the Energy Consumption of Video Streaming: Components, Challenges, ...
Exploring the Energy Consumption of Video Streaming: Components, Challenges, ...
Alpen-Adria-Universität
 
SARENA: SFC-Enabled Architecture for Adaptive Video Streaming Applications
SARENA: SFC-Enabled Architecture for Adaptive Video Streaming ApplicationsSARENA: SFC-Enabled Architecture for Adaptive Video Streaming Applications
SARENA: SFC-Enabled Architecture for Adaptive Video Streaming Applications
Alpen-Adria-Universität
 
LLL-CAdViSE: Live Low-Latency Cloud-based Adaptive Video Streaming Evaluation...
LLL-CAdViSE: Live Low-Latency Cloud-based Adaptive Video Streaming Evaluation...LLL-CAdViSE: Live Low-Latency Cloud-based Adaptive Video Streaming Evaluation...
LLL-CAdViSE: Live Low-Latency Cloud-based Adaptive Video Streaming Evaluation...
Alpen-Adria-Universität
 

More from Alpen-Adria-Universität (20)

VEED: Video Encoding Energy and CO2 Emissions Dataset for AWS EC2 instances
VEED: Video Encoding Energy and CO2 Emissions Dataset for AWS EC2 instancesVEED: Video Encoding Energy and CO2 Emissions Dataset for AWS EC2 instances
VEED: Video Encoding Energy and CO2 Emissions Dataset for AWS EC2 instances
 
GREEM: An Open-Source Energy Measurement Tool for Video Processing
GREEM: An Open-Source Energy Measurement Tool for Video ProcessingGREEM: An Open-Source Energy Measurement Tool for Video Processing
GREEM: An Open-Source Energy Measurement Tool for Video Processing
 
Optimal Quality and Efficiency in Adaptive Live Streaming with JND-Aware Low ...
Optimal Quality and Efficiency in Adaptive Live Streaming with JND-Aware Low ...Optimal Quality and Efficiency in Adaptive Live Streaming with JND-Aware Low ...
Optimal Quality and Efficiency in Adaptive Live Streaming with JND-Aware Low ...
 
VEEP: Video Encoding Energy and CO₂ Emission Prediction
VEEP: Video Encoding Energy and CO₂ Emission PredictionVEEP: Video Encoding Energy and CO₂ Emission Prediction
VEEP: Video Encoding Energy and CO₂ Emission Prediction
 
Content-adaptive Video Coding for HTTP Adaptive Streaming
Content-adaptive Video Coding for HTTP Adaptive StreamingContent-adaptive Video Coding for HTTP Adaptive Streaming
Content-adaptive Video Coding for HTTP Adaptive Streaming
 
Empowerment of Atypical Viewers via Low-Effort Personalized Modeling of Video...
Empowerment of Atypical Viewers via Low-Effort Personalized Modeling of Video...Empowerment of Atypical Viewers via Low-Effort Personalized Modeling of Video...
Empowerment of Atypical Viewers via Low-Effort Personalized Modeling of Video...
 
Empowerment of Atypical Viewers via Low-Effort Personalized Modeling of Vid...
Empowerment of Atypical Viewers  via Low-Effort Personalized Modeling  of Vid...Empowerment of Atypical Viewers  via Low-Effort Personalized Modeling  of Vid...
Empowerment of Atypical Viewers via Low-Effort Personalized Modeling of Vid...
 
Optimizing Video Streaming for Sustainability and Quality: The Role of Prese...
Optimizing Video Streaming  for Sustainability and Quality: The Role of Prese...Optimizing Video Streaming  for Sustainability and Quality: The Role of Prese...
Optimizing Video Streaming for Sustainability and Quality: The Role of Prese...
 
Energy-Efficient Multi-Codec Bitrate-Ladder Estimation for Adaptive Video Str...
Energy-Efficient Multi-Codec Bitrate-Ladder Estimation for Adaptive Video Str...Energy-Efficient Multi-Codec Bitrate-Ladder Estimation for Adaptive Video Str...
Energy-Efficient Multi-Codec Bitrate-Ladder Estimation for Adaptive Video Str...
 
Machine Learning Based Resource Utilization Prediction in the Computing Conti...
Machine Learning Based Resource Utilization Prediction in the Computing Conti...Machine Learning Based Resource Utilization Prediction in the Computing Conti...
Machine Learning Based Resource Utilization Prediction in the Computing Conti...
 
Evaluation of Quality of Experience of ABR Schemes in Gaming Stream
Evaluation of Quality of Experience of ABR Schemes in Gaming StreamEvaluation of Quality of Experience of ABR Schemes in Gaming Stream
Evaluation of Quality of Experience of ABR Schemes in Gaming Stream
 
Network-Assisted Delivery of Adaptive Video Streaming Services through CDN, S...
Network-Assisted Delivery of Adaptive Video Streaming Services through CDN, S...Network-Assisted Delivery of Adaptive Video Streaming Services through CDN, S...
Network-Assisted Delivery of Adaptive Video Streaming Services through CDN, S...
 
Multi-access Edge Computing for Adaptive Video Streaming
Multi-access Edge Computing for Adaptive Video StreamingMulti-access Edge Computing for Adaptive Video Streaming
Multi-access Edge Computing for Adaptive Video Streaming
 
VE-Match: Video Encoding Matching-based Model for Cloud and Edge Computing In...
VE-Match: Video Encoding Matching-based Model for Cloud and Edge Computing In...VE-Match: Video Encoding Matching-based Model for Cloud and Edge Computing In...
VE-Match: Video Encoding Matching-based Model for Cloud and Edge Computing In...
 
Energy Consumption in Video Streaming: Components, Measurements, and Strategies
Energy Consumption in Video Streaming: Components, Measurements, and StrategiesEnergy Consumption in Video Streaming: Components, Measurements, and Strategies
Energy Consumption in Video Streaming: Components, Measurements, and Strategies
 
Exploring the Energy Consumption of Video Streaming: Components, Challenges, ...
Exploring the Energy Consumption of Video Streaming: Components, Challenges, ...Exploring the Energy Consumption of Video Streaming: Components, Challenges, ...
Exploring the Energy Consumption of Video Streaming: Components, Challenges, ...
 
Video Coding Enhancements for HTTP Adaptive Streaming Using Machine Learning
Video Coding Enhancements for HTTP Adaptive Streaming Using Machine LearningVideo Coding Enhancements for HTTP Adaptive Streaming Using Machine Learning
Video Coding Enhancements for HTTP Adaptive Streaming Using Machine Learning
 
SARENA: SFC-Enabled Architecture for Adaptive Video Streaming Applications
SARENA: SFC-Enabled Architecture for Adaptive Video Streaming ApplicationsSARENA: SFC-Enabled Architecture for Adaptive Video Streaming Applications
SARENA: SFC-Enabled Architecture for Adaptive Video Streaming Applications
 
Immersive Video Delivery: From Omnidirectional Video to Holography
Immersive Video Delivery: From Omnidirectional Video to HolographyImmersive Video Delivery: From Omnidirectional Video to Holography
Immersive Video Delivery: From Omnidirectional Video to Holography
 
LLL-CAdViSE: Live Low-Latency Cloud-based Adaptive Video Streaming Evaluation...
LLL-CAdViSE: Live Low-Latency Cloud-based Adaptive Video Streaming Evaluation...LLL-CAdViSE: Live Low-Latency Cloud-based Adaptive Video Streaming Evaluation...
LLL-CAdViSE: Live Low-Latency Cloud-based Adaptive Video Streaming Evaluation...
 

Recently uploaded

Recently uploaded (20)

WebAssembly is Key to Better LLM Performance
WebAssembly is Key to Better LLM PerformanceWebAssembly is Key to Better LLM Performance
WebAssembly is Key to Better LLM Performance
 
How Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdf
How Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdfHow Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdf
How Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdf
 
Oauth 2.0 Introduction and Flows with MuleSoft
Oauth 2.0 Introduction and Flows with MuleSoftOauth 2.0 Introduction and Flows with MuleSoft
Oauth 2.0 Introduction and Flows with MuleSoft
 
Intro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджераIntro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджера
 
Speed Wins: From Kafka to APIs in Minutes
Speed Wins: From Kafka to APIs in MinutesSpeed Wins: From Kafka to APIs in Minutes
Speed Wins: From Kafka to APIs in Minutes
 
THE BEST IPTV in GERMANY for 2024: IPTVreel
THE BEST IPTV in  GERMANY for 2024: IPTVreelTHE BEST IPTV in  GERMANY for 2024: IPTVreel
THE BEST IPTV in GERMANY for 2024: IPTVreel
 
ECS 2024 Teams Premium - Pretty Secure
ECS 2024   Teams Premium - Pretty SecureECS 2024   Teams Premium - Pretty Secure
ECS 2024 Teams Premium - Pretty Secure
 
IESVE for Early Stage Design and Planning
IESVE for Early Stage Design and PlanningIESVE for Early Stage Design and Planning
IESVE for Early Stage Design and Planning
 
Measures in SQL (a talk at SF Distributed Systems meetup, 2024-05-22)
Measures in SQL (a talk at SF Distributed Systems meetup, 2024-05-22)Measures in SQL (a talk at SF Distributed Systems meetup, 2024-05-22)
Measures in SQL (a talk at SF Distributed Systems meetup, 2024-05-22)
 
The Metaverse: Are We There Yet?
The  Metaverse:    Are   We  There  Yet?The  Metaverse:    Are   We  There  Yet?
The Metaverse: Are We There Yet?
 
ASRock Industrial FDO Solutions in Action for Industrial Edge AI _ Kenny at A...
ASRock Industrial FDO Solutions in Action for Industrial Edge AI _ Kenny at A...ASRock Industrial FDO Solutions in Action for Industrial Edge AI _ Kenny at A...
ASRock Industrial FDO Solutions in Action for Industrial Edge AI _ Kenny at A...
 
What's New in Teams Calling, Meetings and Devices April 2024
What's New in Teams Calling, Meetings and Devices April 2024What's New in Teams Calling, Meetings and Devices April 2024
What's New in Teams Calling, Meetings and Devices April 2024
 
Demystifying gRPC in .Net by John Staveley
Demystifying gRPC in .Net by John StaveleyDemystifying gRPC in .Net by John Staveley
Demystifying gRPC in .Net by John Staveley
 
Top 10 Symfony Development Companies 2024
Top 10 Symfony Development Companies 2024Top 10 Symfony Development Companies 2024
Top 10 Symfony Development Companies 2024
 
Strategic AI Integration in Engineering Teams
Strategic AI Integration in Engineering TeamsStrategic AI Integration in Engineering Teams
Strategic AI Integration in Engineering Teams
 
AI revolution and Salesforce, Jiří Karpíšek
AI revolution and Salesforce, Jiří KarpíšekAI revolution and Salesforce, Jiří Karpíšek
AI revolution and Salesforce, Jiří Karpíšek
 
WSO2CONMay2024OpenSourceConferenceDebrief.pptx
WSO2CONMay2024OpenSourceConferenceDebrief.pptxWSO2CONMay2024OpenSourceConferenceDebrief.pptx
WSO2CONMay2024OpenSourceConferenceDebrief.pptx
 
Extensible Python: Robustness through Addition - PyCon 2024
Extensible Python: Robustness through Addition - PyCon 2024Extensible Python: Robustness through Addition - PyCon 2024
Extensible Python: Robustness through Addition - PyCon 2024
 
10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka Doktorová10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka Doktorová
 
Unpacking Value Delivery - Agile Oxford Meetup - May 2024.pptx
Unpacking Value Delivery - Agile Oxford Meetup - May 2024.pptxUnpacking Value Delivery - Agile Oxford Meetup - May 2024.pptx
Unpacking Value Delivery - Agile Oxford Meetup - May 2024.pptx
 

Quality Optimization of Live Streaming Services over HTTP with Reinforcement Learning

  • 1. All rights reserved. ©2020 1 Quality Optimization of Live Streaming Services over HTTP with Reinforcement Learning The IEEE Global Communications Conference (GLOBECOM) 7 -11 December 2021, Madrid, Spain Farzad Tashtarian, Reyhane Falanji, Abdelhak Bentaleb, Alireza Erfanian, Peyman S. Mashhadi, Christian Timmerer, Hermann Hellwagner, and Roger Zimmermann Christian Doppler laboratory ATHENA | Klagenfurt University | Austria farzad.tashtarian@aau.at | https://athena.itec.aau.at/
  • 2. All rights reserved. ©2020 ● Introduction ● Proposed approach ● Results ● Conclusion and future work ● Q&A Agenda All rights reserved. ©2020 2
  • 3. All rights reserved. ©2020 Introduction ● Recent years have seen tremendous growth in HTTP adaptive live video traffic over the Internet. ● HTTP Adaptive Streaming (HAS) has become the de-facto solution allowing Over-The-Top(OTT) services to deliver an acceptable Quality of Experience (QoE) ● Dynamic network conditions and diverse request patterns ○ A large overhead and significant increase in time complexity 3
  • 4. 4 HAS Player Video Distribution Network Internet Video Contribution CDN Server CDN Server CDN Server Origin Server ABR Encoder Live Source HAS Player HAS Player
  • 5. 5 HAS Player Video Distribution Network Internet Video Contribution CDN Server CDN Server CDN Server Origin Server ABR Encoder Live Source HAS Player HAS Player
  • 6. 6 HAS Player Video Distribution Network Internet Video Contribution CDN Server CDN Server CDN Server Origin Server ABR Encoder Live Source HAS Player HAS Player HTTP Request for blue segment HTTP Request for red segment
  • 7. 7 HAS Player Video Distribution Network Internet Video Contribution CDN Server CDN Server CDN Server Origin Server ABR Encoder Live Source HAS Player HAS Player HTTP Response for blue segment HTTP Response for red segment
  • 8. 8 Video Distribution Network Internet Video Contribution CDN Server Origin Server ABR Encoder Live Source HAS Player HTTP Response for blue segment HTTP Response for red segment How to increase clients’ QoE by considering : 1- Network Bandwidth between users and CDN server 2- Number of requests for a similar channel and quality 3- Different serving methods: - Fetch from CDN/origin server - Transcoding from higher quality to lower one - Serving with lower quality
  • 9. All rights reserved. ©2020 Problem Definition How do we process and serve numerous clients’ requests, demanding different live channels with various bitrates, at the network edge to elevate the clients’ perceived QoE in a network with adverse bandwidth conditions? 9
  • 10. 10 Video Distribution Network Internet Video Contribution CDN Server Origin Server ABR Encoder Live Source HAS Player Solution: ROPL, ● A learning-based client request management solution at the edge ● leverage the deep reinforcement learning, ● serve requests of concurrent users joining various HTTP-based live video channels RL-based Virtual Reverse Proxy (RVP) HTTP Request for blue segment
  • 11. 11 Video Distribution Network Internet Video Contribution CDN Server Origin Server ABR Encoder Live Source HAS Player Solution: ROPL, ● A learning-based client request management solution at the edge ● leverage the deep reinforcement learning, ● serve requests of concurrent users joining various HTTP-based live video channels RL-based Virtual Reverse Proxy (RVP) HTTP Response for blue segment
  • 12. All rights reserved. ©2020 ROPL Architecture ● The RVP consists of the following modules: ○ service manager(SM), ○ bitrate classifier (BC), ○ bandwidth monitor (BM), ○ partial cache, ○ deep reinforcement learning (DRL)agent. 12
  • 13. All rights reserved. ©2020 Deep Reinforcement Learning (DRL) Agent ● Proximal Policy Optimization (PPO) ● Policy gradient-based algorithm with an actor-critic approach ● Markov Decision Process (MDP) model ● The translation of our scenario into an MDP model is as follows: ○ State Space ○ Action Space ○ Policy ○ Reward Function 13 All rights reserved. ©2020
  • 14. All rights reserved. ©2020 DRL Agent- State Space 14 Observation Space indicates the number of the same requests for bitrate j (considering an ascending order) of the live channel i is the available bandwidth given from the SM module
  • 15. All rights reserved. ©2020 DRL Agent- Action Space 15 is a set of actions that can be performed by the DRL agent A set of selected actions by the DRL agent at time step τ Indicates the selected action for aggregated requests in the jth queue of the ith live channel with number
  • 16. 16 ACT1: fetching the requested segment s with bitrate j directly from the remote CDN/origin server; ACT2: serving segments with bitrate j* demanded by request i* in the same time step, where j* < j and action of i* is ACT#1 ACT3: serving by transcoding segment s from a higher bitrate j* demanded by request i* where ACT#1 is selected for i* ACT4: do nothing DRL Agent- Action Space ...
  • 17. All rights reserved. ©2020 DRL Agent- Policy ● Policy refers to the logic based on which the agent takes action according to a particular observation. ● The policy is optimized by learning from the environment and updates based on reward signals. ● A desired convergence of policy happens when the agent learns to take the most rewarding action with the highest conditional probability, and a policy update does not change the current one. 17
  • 18. All rights reserved. ©2020 DRL Agent- Reward Function 18 ❏ An undeniable role in the DRL ❏ Guides the agent towards the following purpose: serving requests with the lowest cost and the maximum QoE The proposed reward function consists of ● Serving cost ● Violation penalty
  • 19. DRL Agent- Reward Function - Serving Cost 19 Serving cost C1-C4 for applying ACK1-ACK4 respectively. by considering the cost for serving a requested bitrate with a lower one, coefficient α2 is selected the required transcoding time A coefficient regarding the cost of bandwidth for applying ACT#1 A coefficient regarding the cost of computational resources for applying ACT#3, Serving by fetch Serving by lower bitrates Serving by transcoding
  • 20. All rights reserved. ©2020 DRL Agent- Reward Function - Violation Penalty 20 For every action set in a timestep, penalty signals are encountered as many as the number of violations according to the following constraints: ● CON#1: empty request ○ whenever any action other than ACT#4 is taken for an empty request ● CON#2: bandwidth ○ when a fetch action (ACT#1) is taken, but the bandwidth limit is violated. ● CON#3: existing of lower bitrate ○ when the agent decides to serve requests in a queue by a lower bitrate, however ACT#1 has not been taken for a queue that contains request(s) with similar segment number but lower bitrate; ● CON#4: existing of higher bitrate ○ when a queue is supposed to be served by transcoding from a higher bitrate version of the same segment, but no queue with a higher bitrate request has performed a successful fetch action (ACT#1); ● CON#5: non-empty queue ○ a “do nothing” decision for a non-empty queue.
  • 21. 21 DRL Agent- Reward Function the normalized total cost in time step τ the normalized total penalty in time step τ sum of violation in time step τ sum of actions’ cost and violation in time step τ
  • 22. 22 Heuristic Post-Processing Algorithm Prevent applying invalid actions on requests
  • 23. 23 Performance Evaluation We conduct the performance evaluation in two modes: ● Simulation scenarios ○ six scenarios for different numbers of live channels, players, and various amounts of bandwidth ● Real-world scenarios ○ RVP in Python to serve requests received from the goDASH player ○ HTTP origin server on an AWS virtual machine located in the Frankfurt zone ○ CH1(Big Buck Bunny), CH2(Tears of Steel), CH3(Sintel) CH1: {180p@0.25Mbps, 360p@1.1Mbps, 414p@1.82Mbps, 720p@3.0Mbps, 1080p@3.89Mbps} CH2: {216p@0.39Mbps, 360p@1.1Mbps, 414p@1.82Mbps, 720p@2.38Mbps, 1080p@3.89Mbps} CH3: {288p@0.57Mbps, 414p@1.82Mbps, 720p@2.38Mbps, 1080p@3.89Mbps, 1080p@4.3Mbps}
  • 24. 24 Simulation Experiments Frequency of penalties (a) and taken actions (b) for different scenarios using ROPL model in simulation
  • 25. Simulation Experiments ... 25 Resulted reward (a) and loss (b) using ROPL model in simulations
  • 26. Real-world Experiment 26 Achieved average QoE for each channel clients and different total bandwidth configurations using ROPL in real-world experiments.
  • 27. All rights reserved. ©2020 Conclusion and Future Work ● We proposed ROPL, ○ a reinforcement learning approach ○ optimizing HTTP-based live streaming ○ an edge based mechanism ○ trace-driven simulations and a real-world setup ○ outperforms existing heuristic-based approaches in terms of QoE, with a factor up to 3.7× ● Extend ROPL to support multiple RVPs 27
  • 28. All rights reserved. ©2020 2 Quality Optimization of Live Streaming Services over HTTP with Reinforcement Learning The IEEE Global Communications Conference (GLOBECOM) 7 -11 December 2021, Madrid, Spain Farzad Tashtarian, Reyhane Falanji, Abdelhak Bentaleb, Alireza Erfanian, Peyman S. Mashhadi, Christian Timmerer, Hermann Hellwagner, and Roger Zimmermann Christian Doppler laboratory ATHENA | Klagenfurt University | Austria farzad.tashtarian@aau.at | https://athena.itec.aau.at/