Samira Afzal, Narges Mehran, Sandro Linder, Christian Timmerer, and Radu Prodan
Institute of Information Technology (ITEC), Alpen-Adria-Universität Austria
samira.afzal@aau.at | https://athena.itec.aau.at/
VE-Match: Video Encoding Matching-based Model
for Cloud and Edge Computing Instances
1
Agenda
Motivation and main
objectives
VE-Match
Experimental results
Conclusion
Video streaming accounts for 67.60%
current network traffic
2
https://www.bbc.com/future/article/20200305-why-your-internet-habits-are-not-as-clean-as-you-think
1
2
2
https://www.sandvine.com/hubfs/Sandvine_Redesign_2019/Downloads/2023/reports/Sandvine%20GIPR%202023.pdf
Urgent action is needed against climate change and global greenhouse gas
(GHG) emissions
The carbon footprint of Internet data traffic accounts for about 3.7% of
GHG
Motivation
1
2 3
Main Objectives
Computationally intensive
Costly
Time-consuming
Energy intensive
Video encoding is
Minimizing the aggregate requirements of the media and resource
providers
Minimizing cost or minimizing energy consumption
Making a trade-off between energy usage and cost
To select Cloud/Edge Instances for video encoding/transcoding operations,
aiming at:
VE-Match
4
5
VE-Match
Video encoding application consisting of codec, bitrate, and
resolution set for encoding a video segment
VE-Match, a matching-based method to schedule video encoding
applications on both Cloud and Edge resources to optimize costs
and energy consumption
Video Encoding
Matching-based Model
for Cloud and Edge
Computing Instances
(VE-Match)
VE-Match
6
7
VE-Match Architecture Overview
8
Objective Function
To match each encoding application A to an instance I that minimizes 𝑂 (A, I)
based on two independent goals of cost C and energy E on the players’ sides
where 0 < 𝛼 + 𝛽 ≤ 2 define the competition between the
media (defined by 𝛼 on the application side) and
resource (defined by 𝛽 on the instance side) providers’
Evaluation
Application-side cost-
optimized
Instance-side energy-
optimized
Tradeoff
9
10
Experimental Infrastructure
Germany Austria
Cheapest with
highest energy cons.
Most expensive with
Lowest energy cons.
11
Encoding Energy Benchmark
Benchmark
four types of Cloud AWS EC2
instances and one type of Edge
server
500 video encoding applications
4K video resolutions, HEVC
format
The cost of the medium instance is
higher than that of all AWS Cloud
instances
The energy consumption of the
medium instance is lower than all
AWS Cloud
12
VE-Match Experimental Results
77.85% cost reduction
45.42% energy reduction
13
VE-Match CO2 Emission Analysis
80% CO2 emission reduction
Source materials in power production
Conclusions
The proposed VE-Match
employs game-theoretic principles
aims at minimizing video encoding cost and/or energy by selecting the
appropriate instance types of the Cloud and Edge
We evaluated VE-Match in a real computing testbed
varying the number of video applications
across Cloud and Edge Instances
Our experimental results demonstrate that
Video encoding cost reduction by 17%-78% in the cost-optimized scenarios
Video encoding energy consumption reduction by 38%-45% in the energy-
optimized scenarios along with 80% CO2 emission reduction
14
Thank you
Have a
great day
ahead!
Paper link: https://dl.acm.org/doi/10.1145/3593908.3593943
Institute of Information Technology (ITEC) Alpen-Adria-Universität Austria
samira.afzal@aau.at https://itec.aau.at/

VE-Match: Video Encoding Matching-based Model for Cloud and Edge Computing Instances

  • 1.
    Samira Afzal, NargesMehran, Sandro Linder, Christian Timmerer, and Radu Prodan Institute of Information Technology (ITEC), Alpen-Adria-Universität Austria samira.afzal@aau.at | https://athena.itec.aau.at/ VE-Match: Video Encoding Matching-based Model for Cloud and Edge Computing Instances
  • 2.
  • 3.
    Video streaming accountsfor 67.60% current network traffic 2 https://www.bbc.com/future/article/20200305-why-your-internet-habits-are-not-as-clean-as-you-think 1 2 2 https://www.sandvine.com/hubfs/Sandvine_Redesign_2019/Downloads/2023/reports/Sandvine%20GIPR%202023.pdf Urgent action is needed against climate change and global greenhouse gas (GHG) emissions The carbon footprint of Internet data traffic accounts for about 3.7% of GHG Motivation 1
  • 4.
    2 3 Main Objectives Computationallyintensive Costly Time-consuming Energy intensive Video encoding is Minimizing the aggregate requirements of the media and resource providers Minimizing cost or minimizing energy consumption Making a trade-off between energy usage and cost To select Cloud/Edge Instances for video encoding/transcoding operations, aiming at:
  • 5.
  • 6.
  • 7.
    5 VE-Match Video encoding applicationconsisting of codec, bitrate, and resolution set for encoding a video segment VE-Match, a matching-based method to schedule video encoding applications on both Cloud and Edge resources to optimize costs and energy consumption
  • 8.
    Video Encoding Matching-based Model forCloud and Edge Computing Instances (VE-Match) VE-Match 6
  • 9.
  • 10.
    8 Objective Function To matcheach encoding application A to an instance I that minimizes 𝑂 (A, I) based on two independent goals of cost C and energy E on the players’ sides where 0 < 𝛼 + 𝛽 ≤ 2 define the competition between the media (defined by 𝛼 on the application side) and resource (defined by 𝛽 on the instance side) providers’
  • 11.
  • 12.
    10 Experimental Infrastructure Germany Austria Cheapestwith highest energy cons. Most expensive with Lowest energy cons.
  • 13.
    11 Encoding Energy Benchmark Benchmark fourtypes of Cloud AWS EC2 instances and one type of Edge server 500 video encoding applications 4K video resolutions, HEVC format The cost of the medium instance is higher than that of all AWS Cloud instances The energy consumption of the medium instance is lower than all AWS Cloud
  • 14.
    12 VE-Match Experimental Results 77.85%cost reduction 45.42% energy reduction
  • 15.
    13 VE-Match CO2 EmissionAnalysis 80% CO2 emission reduction Source materials in power production
  • 16.
    Conclusions The proposed VE-Match employsgame-theoretic principles aims at minimizing video encoding cost and/or energy by selecting the appropriate instance types of the Cloud and Edge We evaluated VE-Match in a real computing testbed varying the number of video applications across Cloud and Edge Instances Our experimental results demonstrate that Video encoding cost reduction by 17%-78% in the cost-optimized scenarios Video encoding energy consumption reduction by 38%-45% in the energy- optimized scenarios along with 80% CO2 emission reduction 14
  • 17.
    Thank you Have a greatday ahead! Paper link: https://dl.acm.org/doi/10.1145/3593908.3593943 Institute of Information Technology (ITEC) Alpen-Adria-Universität Austria samira.afzal@aau.at https://itec.aau.at/