EADAS: Edge Assisted Adaptation Scheme for HTTP Adaptive Streaming
0 likes•409 views
The document presents EADAS, an edge-assisted adaptation scheme designed to enhance client-based adaptive bitrate (ABR) streaming by improving adaptation decisions through segment prefetching and clustering users by subscription levels. Experimental results demonstrate a notable increase in quality of experience (QoE) for both premium and basic users, with improvements in throughput and reduced stalls across different ABR algorithms. EADAS is lightweight and aims to provide fairness among users while supporting various subscription models.
EADAS: Edge Assisted Adaptation Scheme for HTTP Adaptive Streaming
- 1. All rights reserved. ©2020 All rights reserved. ©2020 EADAS: Edge Assisted Adaptation Scheme for HTTP Adaptive Streaming 1 IEEE 46th Conference on Local Computer Networks (LCN) October 4-7, 2021 Jesús Aguilar Armijo, Christian Timmerer, and Hermann Hellwagner Christian Doppler laboratory ATHENA | Alpen-Adria-Universität Klagenfurt | Austria [email protected] | https://athena.itec.aau.at/
- 2. All rights reserved. ©2020 ● Introduction ● Algorithm ● Segment prefetching ● Clustering per subscription ● Results ● Q & A Table of content All rights reserved. ©2020 2
- 3. All rights reserved. ©2020 ● Client-based algorithm has limited information available to perform its decisions ● Usually, edge-based ABR algorithms are based on an optimization model with time-slots, where they collect all the requests from the users ○ But requests are not synchronized and they might have different segment duration ● We propose EADAS, an edge-based scheme that supports the client-based ABR algorithm, improving its adaptation decisions ● Provide awareness of other users requests, segment prefetching support and different level of subscription ● Operates in an on-the-fly manner with minimum latency added. It is lightweight in contrast to optimization-based, state-of-the-art time-slotted approaches. Introduction All rights reserved. ©2020 3
- 4. All rights reserved. ©2020 ● EADAS algorithm is executed for each segment request ● It focus on improve QoE and fairness among the users ● The 𝛼 value in our algorithm can prioritize QoE or fairness according to our preferences ● Lower 𝛼 values prioritize fairness, higher alpha values prioritize QoE: final score = 𝛼 x quality score + (1 - 𝛼 ) x fairness score EADAS algorithm All rights reserved. ©2020 4
- 5. All rights reserved. ©2020 ● We study different segment prefetching policies, analyzing costs and benefits ○ Last segment quality (LSQ) ○ Last segment quality plus (LSQ+) ○ All segment qualities (ASQ) ● We test SARA ABR algorithm with different prefetching policies ● Results show that throughput-based or hybrid ABR algorithms are not prepared to support segment prefetching, we have radio throughput miscalculations ● EADAS was designed to support segment prefetching and leverage its benefits EADAS segment prefetching All rights reserved. ©2020 5
- 6. All rights reserved. ©2020 ● Service providers may want to offer different levels of subscriptions to offer several pricing schemes (e.g., basic, premium) to customers with differentiated services, e.g., in terms of QoE ● For example, premium clients may benefit from better segment prefetching policies ● EADAS algorithm can group users with the same characteristics and assure fairness among them ● We conduct experiment with and without EADAS, with half of the clients assign to be premium with segment prefetching LSQ+: ● Results show how EADAS clustering per subscription increase the premium user QoE a 26% (from 3.35 to 4.22) and the basic user QoE a 20% (from 3.45 to 4.14) ● EADAS also increases the fairness among users of the same cluster EADAS clustering per subscription All rights reserved. ©2020 6
- 7. All rights reserved. ©2020 ● As EADAS aims to improve client-based ABR algorithms, we test our mechanism using real 4G radio traces using three client-based ABR algorithms with different approaches: ○ Throughput-based ABR (TBA) ○ Buffer-based ABR (BBA) ○ Hybrid-based ABR (SARA) ● EADAS improves the performance of the three ABR algorithms, improving the mean bitrate and/or reducing the number of stalls ● EADAS improves the QoE by 4.6%, 23.5%, and 24.4% and the mean fairness index by 11%, 3.4% and 5.8% for BBA, TBA, and SARA, respectively EADAS results All rights reserved. ©2020 7
- 8. Thank you Q&A All rights reserved. ©2020 8