You can deploy sample fault tolerant setup for a live streaming pipeline that demonstrates fault resilience features. You can deploy a fault tolerant setup in various ways. For example, you can include Origin and Packaging functionality in a single data center. Alternatively, you can host Origin servers in the cloud, use multiple Packaging sites, or eliminate packagers and use Encoders to package the content.
For the sample setup, you can include two redundantly deployed data centers for Origin service and a single data center for Packaging. Each Origin data center should have two redundantly deployed origin servers, multiple reverse proxies and a load balancer. The packaging center can contain two redundantly deployed packagers and one or more encoders.
The sample deployment scenario has the following features:
Encoders can be configured to publish the stream in MPEG2TS format or through RTMP to the packagers. If the encoders publish the stream on a multicast port in MPEG2TS format, the encoders and packagers must be in the same subnet. However if the encoder publishes RTMP stream, the encoders and packagers can be in separate subnets.
Encoder(s) are configured to generate multiple bitrates of each ingested stream.
For each bitrate, you configure a single stream module on both the packagers.
For each bitrate, you can configure a single stream module on each origin. This module receives the packaged content from both the packagers.
This setup requires the following public IP addresses:
GTM IP address
IP of the load balancer of Origin Data Center 1
IP of the load balancer of Origin Data Center 2
Top-level manifest requests are routed through GTM IP. All the rendition stream level manifest requests are served directly through the Load Balancer IP addresses.
You can configure GTM to load balance between origin data centers where set-level manifest requests are routed to different sites in a fixed ratio, for example 50:50.
If you designate data-center2 as a backup for the primary data center in case of a failure, configure GTM to route the requests in the ratio 100:0. Based on liveness checks, when GTM determines that data-center1 is not responsive, it can switch-over to the backup site by redirecting all new set-level manifests to it. The existing clients fail-over to the URLs of the data-center2 when the primary data center cannot fulfill stream level manifest requests.