A New Integration Opens the Way in the Cloud: Meet Amazon KVS Adapters

A New Integration Opens the Way in the Cloud: Meet Amazon KVS Adapters

We finally merged two new adapters in the Savant Framework, which help integrate Amazon Kinesis Video Streams (KVS) with Savant. KVS is a great technology that combines video-optimized streaming and video storage.

Many AWS AI technologies integrate with KVS, as depicted in the following image:

Amazon KVS architecture overview

They also enable simplified WebRTC application development:

Amazon KVS integration diagram

The adapters are designed to benefit from KVS properties. A commercial customer decided to sponsor adapter development and grant the adapters to the Savant community under the Apache 2.0 license.

We implemented two: KVS Source, which consumes data from KVS streams, and KVS Sink, which delivers data to KVS. Now, you can use KVS capabilities in your hybrid edge-cloud pipelines with them.

The KVS Sink Adapter is a regular adapter that delivers video data to KVS. Because of KVS’s limitations, only H.264 and HEVC codecs are supported. It can automatically create new KVS streams for appearing sources, so you do not need to worry about it.

The adapter differs from the original AWS KVS sink because it supports backpressure, allowing you not to lose video data when the internet connection is unavailable for extended periods. Just combine the adapter with the Buffer Adapter so the video is buffered at the edge of the device when the device is offline. It is especially beneficial for transportation and remote sites with unstable network connections.

The KVS Source Adapter implements both start-up parameters and can be dynamically reconfigured with RESTful API without reloads. It supports playing from a particular time position, stopping, switching to another stream, etc. These features are massive because they allow for implementing on-demand processing or playback.

Read more on how to use them in the documentation:

Do not hesitate to contact us on Discord to learn more about them. We are going to publish an end-to-end sample.