AWS for M&E Blog

Nippon Television Network Corporation implements cloud playout using AWS

A playout system is used to stream video for broadcasting and over-the-top (OTT) video. It performs tasks such as: switching between a live feed and a pre-recorded asset; controlling and editing the progress of a TV program; and switching video sources if there is an emergency. Frame-level accuracy is vital because switching between a live feed and an asset requires the user to ensure that only a specific video segment is processed. The same is true when a video needs to be “masked”, i.e. another video must be overlaid on it if the video cannot be broadcast due to a copyright issue. The technology to accomplish all of these tasks used to require dedicated hardware and several years to build and plan, resulting in substantial cost.

In October 2020, Nippon Television Network Corporation (NTV) started a live streaming trial (NTV live streaming trial) of prime-time shows from terrestrial TV channels. In preparation for the trial, NTV adopted a cloud playout service named “PLAY KRONOS (KRONOS)” by PLAY Co., Ltd instead of a traditional playout system. KRONOS is a cloud playout service that as part of its video streaming backend leverages Nimbus, a Veset linear channel management platform, and uses AWS services such as AWS Media Services and Amazon Elastic Container Service (Amazon ECS) to achieve high availability.

Gaku Matsumoto, Vice Manager of Engineering Division, ICT Strategy Headquarters, Nippon Television said: “By adopting KRONOS, we were able to implement NTV’s live streaming in just five months, a third of the time it would normally take and without procuring any dedicated hardware. In addition, since KRONOS is available as SaaS employing a pay-as-you-go pricing model, we found it easy to introduce the system and were able to reduce the initial building cost. We started the NTV live streaming as a trial, to get data on many different aspects such as viewer reaction, whether it would work seamlessly with other systems, and how we could establish the operational workflow. We are glad that we were able to launch the service quickly at low cost.”

Cloud playout workflow

the cloud playout workflow from the datacenter to the end user

The cloud playout workflow from the datacenter to the end user

The following is a detailed look at the architecture of the cloud playout system, which is designed for simultaneous online streaming:

  • The signal is output from the encoder at the central apparatus room of the NTV, and sent via AWS Direct Connect to AWS Elemental MediaConnect.
  • MediaConnect then outputs the signal to a playout service (Amazon EC2 was used in this workflow)
  • The playout service switches between the live feed and pre-recorded video (fetched from Amazon Simple Storage Service), which is triggered by an SCTE signal, a specified time (system clock), or the end of the previous event.
  • An APC service (Amazon EC2 in this workflow) controls and manages the schedule for switching between the asset and live feed according to the order from the dashboard service and API Service built using AWS Fargate.
  • The playout service coordinates with APC Service to handle events such as emergency switching.
  • The signal is output from the playout service to MediaConnect, and then it is output from MediaConnect and distributed to a streaming platform named TVer and a broadcast data sharing system named Traffic Sim RecShare CLOUD.
  • The content is then delivered from TVer to the viewer.

Switching scenes with frame-level accuracy

In playout, when switching between the live feed and a pre-recorded asset, frame-level accuracy is required for a seamless switch. That means switching must be performed at a granularity smaller than a second or two (one frame is equivalent to about 0.0334 seconds at 29.97 frames per second). Identifying a segment and processing it precisely with frame-level accuracy was not feasible by only referring to the system clock running in the server. To achieve frame-level accuracy for the NTV live streaming trial, they employed a method to switch between different input sources based on SCTE35 splice_insert (KRONOS also supports SCTE35 time signal) superimposed on the live feed. SCTE35 (splice_insert/time signal) is a standard to signal an advertisement insertion and program start/end. Using an ad insertion marker, they achieved frame-level accuracy and an advertisement insertion for the NTV live streaming trial.

Daisuke Miyajima, COO of PLAY, the provider of KRONOS, says: “Through our close R&D collaboration with Veset we developed the input switching mechanism with frame-level accuracy. In addition, KRONOS incorporates managed services such as MediaConnect and Fargate. Leveraging managed services has enabled us to reduce person-hours for system operations and management, and to focus on developing features requested by our customers. We are still receiving many feature requests, and KRONOS will continue to evolve.”

Leveraging cloud playout to quickly respond to new service requirements

Matsumoto adds: “For a broadcaster like us, KRONOS — built on AWS and provided as SaaS — is of great value in that the development cost is not borne solely by a single company and there is no need to own a fixed asset. KRONOS also allows us to control the progress of a TV program and edit it by accessing via the web application in a secure way. This allows for operational flexibility. NTV can even enable a staff member of a different TV station to edit a TV program if it was originally produced by them. Service requirements for streaming, unlike those for broadcasting, change rapidly as the environment changes. We believe that building the core playout part on AWS has led to more flexibility in system configuration, enabling us to respond to new requirements in a timely manner.”

NTV Live Streaming Trial
Cloud Playout service「 PLAY KRONOS 」