# How MediaPackage works


AWS Elemental MediaPackage uses just-in-time format conversion to deliver over-the-top (OTT) video from a single source to a wide variety of playback devices or content delivery networks (CDNs).

In the processing flow for live content, encoders send live HLS streams to MediaPackage. MediaPackage then packages the content, formatting it in response to playback requests from downstream devices. 

The following sections describe the live processing flows.

**Topics**
+ [

# General AWS Elemental MediaPackage live processing flow
](what-is-flow-gen.md)
+ [

# Live input redundancy AWS Elemental MediaPackage processing flow
](what-is-flow-ir.md)

# General AWS Elemental MediaPackage live processing flow
General live processing flow

The following outlines the general flow of live content in MediaPackage:

1. An upstream encoder (such as AWS Elemental MediaLive) sends an HLS live stream using AWS Signature Version 4 to authorize request to your origin and your IAM channel policy. If you're using input redundancy, the encoder sends two identical HLS live streams to MediaPackage, one to each ingest domain on the channel. MediaPackage uses the stream from one ingest URL as the source content. If MediaPackage stops receiving content on the active ingest URL, it automatically switches to the other ingest URL for source content. Additionally, AWS scales resources up and down to handle the incoming traffic.

   For more information, see [Live input redundancy AWS Elemental MediaPackage processing flow](what-is-flow-ir.md).
**Note**  
To permit support for features like time-shifted viewing, MediaPackage stores all received content for a limited time. This stored content is only available for playback if it falls within the **startover window** that's defined on the endpoint. Stored content isn't available for playback if it's outside the startover window, or if you haven't defined a window on the endpoint. For more information, see [Time-shifted viewing with AWS Elemental MediaPackage](time-shifted.md).

1. A downstream device requests content from MediaPackage through the endpoint egress domain. A downstream device is either a video player or a CDN. The egress domain is associated with a channel group and an endpoint for a specific streaming format (either TS or CMAF).

1. When MediaPackage receives the playback request from the downstream device, it dynamically packages the stream according to the settings that you specified on the origin endpoint. Packaging can include adding encryption and configuring audio, video, and subtitles or captions track outputs.

   Be sure to order your inputs so that your preferred audio rendition is listed first in the audio section of the multivariant playlist. Do the same for the subtitles or captions. When packaging audio and subtitles or captions tracks, MediaPackage designates the first audio and captions or subtitles track as `DEFAULT=YES` and `AUTO-SELECT=YES`. This packaging overrides default and auto-select settings from the input.

1. MediaPackage delivers the output stream over HTTPS to the requesting device. As with input, AWS scales resources up and down to handle changes in traffic.

Throughout the content input and output processes, MediaPackage detects and mitigates potential infrastructure failures before they become a problem for viewers. 

The following illustration shows the overall process.

![\[MediaPackage workflow from encoder to MediaPackage packaging content to downstream device.\]](http://docs.aws.amazon.com/mediapackage/latest/userguide/images/bbl flow2-empv2.png)


# Live input redundancy AWS Elemental MediaPackage processing flow
Live input redundancy processing flow

Achieve input redundancy in AWS Elemental MediaPackage by sending two streams to separate ingest domains on a channel in MediaPackage. One of the streams becomes the primary, active source of content for the endpoints, while the other continues to passively receive content. If MediaPackage stops receiving content from the active stream, it switches over to the other ingest stream so that content playback isn't interrupted.

If you use MediaPackage with AWS Elemental MediaLive (for example), here's the flow of input redundancy:

1. You create a channel group in MediaPackage, as described in [Creating a channel group in AWS Elemental MediaPackage](channel-group-create.md). When MediaPackage provisions the channel group, it creates an egress domain for all channels and origin endpoints within the channel group.

1. You create a channel within the channel group as described in [Creating a channel in AWS Elemental MediaPackageCreating a channel](channels-create.md). When MediaPackage provisions the channel, it creates two ingest domains for the channel. If you're not using input redundancy, you can send a stream to either ingest domain. There's no requirement that you send content to both domains.

1. You create an origin endpoint within the channel as described in [Creating an origin endpoint in AWS Elemental MediaPackage](endpoints-create.md). 
**Important**  
If you use short output segments, depending on your playback device, you might see buffering when MediaPackage switches inputs. You can reduce buffering by using the time delay feature on the endpoint. Be aware that using a time delay introduces latency to end-to-end delivery of the content. For information about enabling time delay, see [Creating an origin endpoint in AWS Elemental MediaPackage](endpoints-create.md).

1. You create an input and channel in AWS Elemental MediaLive, and you add a MediaPackage output group to the channel in MediaLive. For more information, see [Creating a Channel from Scratch](https://docs.aws.amazon.com/medialive/latest/ug/creating-channel-scratch.html) in the *AWS Elemental MediaLive User Guide*. 

   If you use an HLS output group in AWS Elemental MediaLive, the input loss action on the HLS group's settings must be set to pause the output if the service doesn't receive input. If MediaLive sends a black frame or some other filler frame when it's missing input, then MediaPackage can't tell when segments are missing, and subsequently can't perform failover. For more information about setting the input loss action in MediaLive, see [Fields for the HLS Group](https://docs.aws.amazon.com/medialive/latest/ug/hls-group-fields.html) in the *AWS Elemental MediaLive User Guide*. 
**Important**  
If you use a different encoder (not AWS Elemental MediaLive) and you send two separate streams to the same channel in MediaPackage, the streams must have identical encoder settings and manifest names. Otherwise, input redundancy might not work correctly and playback could be interrupted if the inputs switch.

1. You start the channel in AWS Elemental MediaLive to send the streams to MediaPackage.

1. MediaPackage receives content on both of the ingest URLs, but only one of the streams is used for source content at a time. If the active stream is missing any segments, then MediaPackage automatically fails over to the other stream. MediaPackage continues to use this stream until failover is needed again.

   The formula that's used to determine if an input is missing segments is based on the segment lengths on the inputs and the endpoints. If an input is missing segments and quickly recovers, an endpoint with longer segment lengths won't switch inputs. This might result in different endpoints on the channel using different inputs (if one endpoint switches and the other doesn't). This is expected behavior and should not affect the content workflow.