Columbia

Architecture and Design

Lab

vbench: a Benchmark for Video Transcoding in the Cloud

Overview

vbench is the first benchmark for the emerging video-as-a-service workload and will be presented at ASPLOS 2018 [1].

Current estimates place video at 71% of evening traffic in North America and project that this number will grow to 80% by 2020. Each video that is streamed over the internet has been transcoded at least once by a cloud video provider such as YouTube or Facebook. This creates a great computational burden on the infrastructure of these companies. Furthermore, the rate of growth of uploaded video content has now surpassed the rate of growth of computational power for standard CMP systems. New solutions are necessary in this space.

vbench captures the nuances of this important workload including the tradeoffs that cloud video providers have to navigate between Quality of Experience (QoE), network cost, compute cost, and storage cost.

Download

vbench is based on open source software and freely redistributable video sequences.

We would like to keep track of who is using vbench so please complete the following form to download the benchmark.


If you have questions on the usage of the benchmark contact ude.aibmuloc.sc.stsil@hcnebv

Support

This material is based upon work supported by 2016 Google Faculty Research Award, NSF (CCF-1065338) and C-FAR, one of six centers of STARnet, a Semiconductor Research Corporation program sponsored by MARCO and DARPA. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of these sponsors.

Publications

Please cite this summary paper when you publish results obtained using vbench:

[1] Andrea Lottarini, Alex Ramirez, Joel Coburn, Martha A. Kim, Parthasarathy Ranganathan, Daniel Stodolsky, Mark Wachsler. vbench: Benchmarking Video Transcoding in the Cloud. In Proceedings of the Twenty-third International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), ASPLOS ’18, New York, NY, USA, 2018. ACM. [Download]