Encoding and transcoding videos into multiple codecs and representations is a significant challenge that requires seconds or even days on high-performance computers depending on many technical characteristics, such as video complexity or encoding parameters. Cloud computing offering on-demand computing resources optimized to meet the needs of customers and their budgets is a promising technology for accelerating dynamic transcoding workloads. In this work, we propose OTEC, a novel multi-objective optimization method based on the mixed-integer linear programming model to optimize the computing instance selection for transcoding processes. OTEC determines the type and number of cloud and fog resource instances for video encoding and transcoding tasks with optimized computation cost and time. We evaluated OTEC on AWS EC2 and Exoscale instances for various administrator priorities, the number of encoded video segments, and segment transcoding times. The results show that OTEC can achieve appropriate resource selections and satisfy the administrator’s priorities in terms of time and cost minimization.
OTEC: An Optimized Transcoding Task Scheduler for Cloud and Fog Environments
1. Samira Afzal, Farzad Tashtarian, Hamid Hadian, Alireza Erfanian, Christian Timmerer, and Radu Prodan
Institute of Information Technology (ITEC), Alpen-Adria-Universität Austria
samira.afzal@aau.at | https://athena.itec.aau.at/
OTEC: An Optimized Transcoding Task Scheduler for
Cloud and Fog Environments
APOLLO
3. Motivation and Main Objectives
Video streaming is a significant part of the current network traffic
Computationally intensive
Costly
Time-consuming
Video encoding is
Minimizing cost and/or minimizing encoding time
Making a trade-off between encoding time and cost
Considering encoding time deadline and cost limit
To select Cloud/Fog resources for video encoding/transcoding operations,
aiming at:
2
8. Optimization Model
Cluster selection
One cluster per timeslot
Select the required number of instances
Cluster start and end times
Non-overlapping encoding timeslots
Time and cost
Total transcoding time
Total transcoding cost
Sets of constraints:
8
9. Objective Function
9
To optimize a multi-objective function representing a
linear combination of two metrics, total transcoding cost 𝜆
and total transcoding time 𝜃
10. Maximum possible cost
10
Objective Function
To optimize a multi-objective function representing a
linear combination of two metrics, total transcoding cost 𝜆
and total transcoding time 𝜃
Administration priority
𝛼 ∈ [0, 1]
Time-optimized (𝛼 = 0)
Cost-optimized (𝛼 = 1)
Maximum possible
transcoding time
13. Administration Priority
Time-optimized results are 30.5% faster and 24.78% more expensive than the cost-optimized ones.
For 𝛼 = 0.5, there is a tradeoff between the total cost (0.0024 $/s) and transcoding time (6.99 s).
13
15. Timeslot Duration and Number of Segments
Time-optimized scenario
Coordinator on a local cloud instance
A four-core Intel core-i7 processor
32 GB of memory
370 instances
15
16. Timeslot Duration and Number of Segments
Time-optimized scenario
Coordinator on a local cloud instance
A four-core Intel core-i7 processor
32 GB of memory
370 instances
16
17. Timeslot Duration and Number of Segments
Time-optimized scenario
Coordinator on a local cloud instance
A four-core Intel core-i7 processor
32 GB of memory
370 instances
17
18. Timeslot Duration and Number of Segments
∼2.5 times
15%
Time-optimized scenario
Coordinator on a local cloud instance
A four-core Intel core-i7 processor
32 GB of memory
370 instances
18
19. Timeslot Duration and Number of Segments
∼5.24 times
Time-optimized scenario
Coordinator on a local cloud instance
A four-core Intel core-i7 processor
32 GB of memory
370 instances
19
20. Coordinator
4.5 times
8.31 times
Segment Transcoding Time
Time-optimized scenario
Coordinator on a local cloud instance
A four-core Intel core-i7 processor
32 GB of memory
370 instances
10 segments
20
21. Conclusions
Proposed OTEC for optimized scheduling of adaptive transcoding processes in
Cloud and Fog environments
Proposed an MILP optimization model
Highlighted factors:
Administrator priorities, such as cost and/or time priorities
Resources priorities, location, type, cost, number of instances for each
resource type
Segment properties, such as duration, number of segments, segment
transcoding time
Evaluated OTEC on a set of Cloud AWS and Fog Exoscale resource instances
Achieved that time-optimized scenarios are 30.5% faster and 24.78% more
expensive than the cost-optimized ones.
21
22. Thank you
Have a
great day
ahead!
Institute of Information Technology (ITEC) Alpen-Adria-Universität Austria
samira.afzal@aau.at https://itec.aau.at/
APOLLO