You can deploy multiple servers behind a load balancer to distribute the client load over multiple servers. Deploying multiple servers enables you to scale an application for more clients and creates redundancy, which eliminates single points of failure. You can deploy any version of the server (Adobe Media Interactive Server, Adobe Media Development Server, or Adobe Media Streaming Server) in a cluster.
Add all the servers in the cluster to the pool (server farm) in the load balancer. The load balancer distributes traffic among all the servers in the pool. Configure the load balancer to distribute the load in round-robin mode and to monitor over TCP port 1935.
If the server does not have an externally visible IP address, then for HTTP tunnelling to work, you should enable cookies when you deploy servers behind a load balancer. The load balancer checks the cookie and sends requests with this cookie to the same server. Cookies can be enabled in the load balancer or in the Adaptor.xml configuration file in the Adaptor/HTTPTunnel/SetCookie element.
For tunnelling connections, cookies are currently supported only on Adobe® Flash® Player 9.0.28 or later in Windows only.
Distribute introductions across a server-only RTMFP NetGroup.
This technique is new in Adobe Media Server 5. Use a Server-Side ActionScript API to distribute peer lookup requests across multiple servers. For more information, see Distributing introductions across multiple servers.
Set up an edge-origin configuration.
In the Vhost.xml configuration file on each edge server, set <Mode>remote</Mode>. For more information, see Configure edge servers.
On the client, the NetConnection URL changes from "rtmfp://ams-server/application" to "rtmfp://edge-ams-server/?rtmfp://origin-ams-server/application". Alternately, you can configure the RouteTable and RouteEntry in the Vhost.xml file on the edge server to point to the origin server. In this case, the client URL is still "rtmfp://ams-server/application".
Set up DNS entries to distribute the load across the edge servers. For the server to function properly as an RTMFP introducer, use one origin server per application. For peer lookup to work, clients must be connected to the same server process, but not necessarily to the same application. Peer lookups do not work across origin servers. However, peerIDs are cached on the edge server which reduces the load on the origin. When clients are connected to the same edge, the introduction can happen at the edge and there is not a lookup call to the origin.