Skip to main content

Video Miner Architecture

livepeer currently supports 3 modes of operation for video miners:

  • Standalone orchestrator mode (-orchestrator)
  • Standalone transcoder mode (-transcoder)
  • Combined orchestrator and transcoder mode (-orchestrator -transcoder)

An introduction to orchestrators and transcoders can be found in the getting started section. Given these modes of operation, video miners can consider different architectures. Two of the most common architectures are:

  • Combined orchestrator and transcoder
  • Split orchestrator and transcoder

Video miner architectures

The combined orchestrator and transcoder architecture allows the orchestrator to also perform transcoding using its own hardware. Some of the benefits of using this architecture include:

  • Simple to configure
  • Minimizes the number of network hops since transcoding takes place locally on the orchestrator

The split orchestrator and transcoder architecture allows the orchestrator to outsource transcoding to remote transcoders. These transcoders can be run by the same operator as the orchestrator or by third parties. For more information about the latter, see the transcoding pools section. Some of the benefits of using this architecture include:

  • Enables operators to leverage remote hardware to increase an orchestrator's transcoding capacity
  • Allows hardware connected to machines that cannot be made publicly accessible (i.e. lack sufficient IPs, security concerns if the machines also store sensitive data, etc.) to contribute to an orchestrator's transcoding capacity

An important consideration when using the split orchestrator and transcoder architecture is the geographic proximity of the orchestrator and its transcoders. Since orchestrators need to download results from transcoders before returning the results to broadcasters, the extra network hop adds additional latency to the round trip time between a broadcaster and an orchestrator. Ways to minimize the additional latency from this extra network hop include running the transcoders close to the orchestrators and making sure the transcoders have sufficient bandwidth for quick downloads/uploads with the orchestrator at least for the amount of transcoding capacity the transcoders are expected to contribute.