You Get What You Pay For—Except You Don’t: Instance Families and Scaling Strategies for Cloud Video Transcoding

• Breno José Ribeiro de Vasconcelos UFPE
• Elisa de Fátima Andrade Soares UFPE
• Carlos André Guimarães Ferraz UFPE

DOI: https://doi.org/10.5753/webmedia.2026.30601

Resumo

Cloud video transcoding is widely deployed in video-as-aservice pipelines, yet practitioners lack statistically grounded guidance for selecting and scheduling EC2 instance types under real cost constraints. Conventional wisdom recommends compute-optimized instances for encoding workloads, but this ignores provisioning dynamics that can dominate wallclock time and expenditure. Existing benchmarks compare instance families without controlling for provisioning latency or scaling strategy, leaving the true cost driver unidentified. We address this gap through a large-scale controlled experiment spanning more than 5,700 trials across 15 instance types, three codecs (H.264, H.265, VP9), and three scaling strategies, with 30 repetitions per configuration and 95% confidence intervals throughout. Our methodology decouples provisioning from encoding time and evaluates custom-AMI pre-warming, horizontal scaling, and chunk-size sensitivity across resolutions. We show that provisioning latency, not CPU encoding, is the dominant cost factor: naïve horizontal scaling lets provisioning consume up to 93.3% of wall-clock time, whereas a pre-warmed 10-node t3.micro cluster outperforms a single m5.4xlarge by 1.28× at 89.4% lower compute cost. These findings are translated into concrete, statistically grounded guidelines for sizing and scheduling cloud transcoding clusters.

Palavras-chave: Cloud Computing, Video-as-a-Service, Cost Optimization, Video Transcoding, Horizontal Scaling, Provisioning Overhead, Elasticity, ARM Graviton

Referências

P. Álvarez, S. Hernández, J. Fabra, and J. Ezpeleta. 2018. Costdriven provisioning and execution of a computing-intensive service on the Amazon EC2. Comput. J. 61, 9 (2018), 1407–1421. DOI: 10.1093/comjnl/bxy006

O. Alzakholi, L. M. Haji, H. M. Shukur, R. R. Zebari, S. M. Abas, and M. Sadeeq. 2020. Comparison Among Cloud Technologies and Cloud Performance. J. Appl. Sci. Technol. Trends 1, 1 (2020), 40–47. DOI: 10.38094/jastt1219

Amazon Web Services. 2024. AWS Elemental MediaConvert. AWS Documentation. [Online]. Available: [link] [Accessed: Mar. 2026].

Amazon Web Services. 2024. Burstable performance instances. Amazon EC2 User Guide. [Online]. Available: [link] [Accessed: Mar. 2026].

Amazon Web Services. 2024. Warm pools for Amazon EC2 Auto Scaling. Amazon EC2 Auto Scaling User Guide. [Online]. Available: [link] [Accessed: Aug. 2026].

Amazon Web Services. 2026. Amazon EC2 Instance Types. [link]. Accessed Jun. 2026.

L. Araújo, A. Duarte, G. Correa, B. Zatt, and D. Palomino. 2024. Fast ISP Mode Decision for the Versatile Video Coding Intra Prediction Using Machine Learning. In Proc. 30th Brazilian Symp. Multimedia Web. SBC, Juiz de Fora, Brazil, 162–170. DOI: 10.5753/webmedia.2024.243129

O. A. Ben-Yehuda, M. Ben-Yehuda, A. Schuster, and D. Tsafrir. 2013. Deconstructing Amazon EC2 Spot Instance Pricing. ACM Trans. Econ. Comput. 1, 3 (2013), 16:1–16:20. DOI: 10.1145/2509413.2509416

T. Dancheva, U. Alonso, and M. D. Barton. 2023. Cloud benchmarking and performance analysis of an HPC application in Amazon EC2. Cluster Computing 27, 2 (2023), 2273–2290. DOI: 10.1007/s10586-023-04060-4

B. Farley, A. Juels, V. Varadarajan, T. Ristenpart, K. D. Bowers, and M. M. Swift. 2012. More for Your Money: Exploiting Performance Heterogeneity in Public Clouds. In Proc. 3rd ACM Symp. Cloud Comput. ACM, San Jose, CA, USA, 20:1–20:14. DOI: 10.1145/2391229.2391249

W. Felter, A. Ferreira, R. Rajamony, and J. Rubio. 2015. An Updated Performance Comparison of Virtual Machines and Linux Containers. In 2015 IEEE Int. Symp. Perform. Anal. Syst. Softw. (ISPASS). Philadelphia, PA, USA, 171–172. DOI: 10.1109/ISPASS.2015.7095802

S. Fouladi, R. S. Wahby, B. Shacklett, K. V. Balasubramaniam, W. Zeng, R. Bhalerao, A. Sivaraman, G. Porter, and K. Winstein. 2017. Encoding, Fast and Slow: Low-Latency Video Processing Using Thousands of Tiny Threads. In Proc. 14th USENIX Symp. Networked Systems Design and Implementation (NSDI). USENIX Association, Boston, MA, USA, 363–376. [Online]. Available: [link] [Accessed: Mar. 2026].

L. M. Haji, S. R. M. Zeebaree, O. M. Ahmed, A. B. Sallow, K. Jacksi, and R. R. Zeabri. 2020. Dynamic Resource Allocation for Distributed Systems and Cloud Computing. Test Engineering and Management 83, 1 (2020), 22417–22426.

F. Jiang, K. Ferriter, and C. Castillo. 2020. A Cloud-Agnostic Framework to Enable Cost-Aware Scheduling of Applications in a Multi-Cloud Environment. In 2020 IEEE/IFIP Network Operations and Management Symposium (NOMS). IEEE, Budapest, Hungary, 1–9. DOI: 10.1109/NOMS47738.2020.9110325

A. Katsenou, X. Wang, D. Schien, and D. Bull. 2024. Comparative Study of Hardware and Software Power Measurements in Video Compression. In 2024 Picture Coding Symp. (PCS). IEEE, Taichung, Taiwan, 1–5. DOI: 10.1109/PCS60826.2024.10566286

W. Maas, F. Rossi, M. Caggiani, P. Navaux, and A. Lorenzon. 2025. Investigating cloud instances to achieve optimal trade-offs between performance-cost efficiency. Computing 107, 1 (2025), 82–97. DOI: 10.1007/s00607-025-01444-9

W. Maas, F. D. Rossi, M. C. Luizelli, P. O. A. Navaux, and A. F. Lorenzon. 2025. Towards Optimizing Cost and Performance for Parallel Workloads in Cloud Computing. In Proc. 15th Int. Conf. Cloud Comput. and Services Science (CLOSER). SciTePress, Porto, Portugal, 231–238. DOI: 10.5220/0013421600003950

A. Mahida. 2022. Comprehensive Review on Optimizing Resource Allocation in Cloud Computing for Cost Efficiency. J. Artif. Intell. Cloud Comput. 1, 2 (2022), 1–4. DOI: 10.47363/JAICC/2022(1)232

W. Moina-Rivera, M. Garcia-Pineda, J. Gutiérrez-Aguado, and J. M. Alcaraz-Calero. 2024. Cloud media video encoding: review and challenges. Multimedia Tools and Applications 83 (2024), 81231–81278. DOI: 10.1007/s11042-024-18763-2

MulticoreWare Inc. 2023. x265 HEVC Encoder Documentation: Architecture-specific Optimizations. x265 Project Documentation. [Online]. Available: [link] [Accessed: Mar. 2026].

T. S. Nikoui, A. Balador, A. M. Rahmani, and Z. Bakhshi. 2020. Cost-Aware Task Scheduling in Fog-Cloud Environment. In 2020 CSI/CPSSI Int. Symp. Real-Time Embedded Syst. Technol. (RTEST). IEEE, Tehran, Iran, 1–8. DOI: 10.1109/RTEST49666.2020.9140118

Z. Ou, H. Zhuang, A. Lukyanenko, J. K. Nurminen, P. Hui, V. V. Mazalov, and A. Ylä-Jääski. 2013. Is the Same Instance Type Created Equal? Exploiting Heterogeneity of Public Clouds. IEEE Trans. Cloud Comput. 1, 2 (2013), 201–214. DOI: 10.1109/TCC.2013.12

Z. Ou, H. Zhuang, J. K. Nurminen, A. Ylä-Jääski, and P. Hui. 2012. Exploiting Hardware Heterogeneity within the Same Instance Type of Amazon EC2. In Proc. 4th USENIX Conf. Hot Topics Cloud Comput. USENIX Association, Boston, MA, USA, 4–4.

R. Rosado, O. Santos, F. Oliveira, L. Silva, G. Sanchez, I. Storch, D. Palomino, and L. Agostini. 2024. A Coding-Efficiency-Aware Fast Heuristic for VVC Intra-Frame Prediction Targeting 360° Videos. In Proc. 30th Brazilian Symp. Multimedia Web. SBC, Juiz de Fora, Brazil, 355–359. DOI: 10.5753/webmedia.2024.243128

J. Scheuner and P. Leitner. 2018. Estimating Cloud Application Performance Based on Micro-Benchmark Profiling. In 2018 IEEE 11th Int. Conf. Cloud Comput. (CLOUD). IEEE, San Francisco, CA, USA, 90–97. DOI: 10.1109/CLOUD.2018.00019

Y. Yang and H. Casanova. 2003. RUMR: Robust Scheduling for Divisible Workloads. In Proc. 12th IEEE Int. Symp. High Perform. Distrib. Comput. IEEE, Seattle, WA, USA, 114–123. DOI: 10.1109/HPDC.2003.1210021

A. Zabrovskiy, P. Agrawal, V. Kashansky, R. Kersche, C. Timmerer, and R. Prodan. 2022. FSpot: Fast and Efficient Video Encoding Workloads Over Amazon Spot Instances. Computers, Materials & Continua 71, 3 (2022), 5677–5694. DOI: 10.32604/cmc.2022.023630