Scenario: Distributed streaming media caches

About this scenario

The managers of NetDevices, Inc., want to provide live training sessions to their employees in offices around the world. They want to deploy streaming media service transparently, with L4 switches deployed with streaming media caches.

Organization's requirements

The organization's requirements are as follows:

Use bandwidth efficiently
- The planner wants a deployment scheme that uses bandwidth as efficiently as possible.
  Bandwidth is quite expensive in some of the countries in which the company's offices are located. The planner wants to avoid purchasing additional bandwidth or, if that is not possible, to minimize the amount of bandwidth that must be purchased.
- The planner anticipates that the company's current bandwidth cannot support the traffic that would result if clients from all branch offices sent requests directly to the streaming server in the U.S.A. The planner wants to be able to deploy streaming media service so that most requests from the European regions are resolved without sending the requests to the U.S.A.
Protect the streaming server from becoming overloaded
The planner expects that the streaming server in the U.S.A. cannot handle the number of connections that would result if all clients connect directly to the server. The planner, therefore, needs a solution that "protects" the streaming server.
Maintain the same quality of service for nonstreaming traffic as that was provided before deploying streaming media service
The planner wants to be sure that the increase in traffic resulting from streaming media service does not adversely affect the quality of service for other types of network traffic.
Provide good quality streaming media

Deployment illustrated

The following illustration shows how streaming media caches were deployed at NetDevices, Inc., and their relationship with the streaming server. The streaming media caches are deployed in regions to be able to serve streams closer to the users.

Note
For simplicity, network devices other than the L4 switches, streaming servers, and streaming media caches are not shown in the illustration.

The arrows in the previous illustration show, for each streaming media cache, where a streaming request is sent next if the streaming media cache in the region is not already delivering the stream to a client (that is, if a streaming miss occurs).

Deployment of transparency with L4 switches

An L4 switch was deployed in each region in front of the streaming media cache. Each L4 switch is set up for transparent proxying, configured to pass requests for port 1755 (MMS) and port 554 (RTSP) to the streaming media cache. (In this example, the organization supports streaming media for both MMS and RTSP.) The switch is also configured to pass HTTP (port 80) traffic to the streaming media caches to support client media browsers that are set for HTTP transport only.

Note
A WCCP router could have been deployed in front of the streaming media caches instead of an L4 switch.

Hierarchical routing

In this organization, routing is set up in a hierarchical fashion. Traffic of any type from the German and French regions destined for the U.S.A. must be sent through the European region.

Note
Alternatively, you can create a logical NetCache hierarchy to distribute requests that a NetCache appliance cannot resolve to one or more a proxy-cache servers higher in the logical hierarchy. See the Administration Guide for details about hierarchies.

Request resolution

Only a subset of the company's streaming requests must be sent over the Internet to the streaming server because the regional streaming media caches attempt to resolve streaming requests that they receive.

Request resolution by the French and German streaming media caches

If the French or German streaming media cache receives a request for a live stream that it is already delivering to a client, it splits the stream for any additional clients that send requests for the same live stream.
If the French or German streaming media cache is not already delivering a stream that a client requests, the cache sends the request to the next level up in the hierarchy--to the European streaming media cache. For this reason, the hierarchical routing scheme works especially well for the streaming media deployment in Europe.

Streaming misses sent to the European region

If the European streaming media cache is already delivering a requested live stream to the German cache and the French cache sends a request for the same stream, the European cache splits the live stream and delivers the stream to the French cache. The same is true if the German cache sends a streaming request to the European cache and the European cache is already delivering the stream to the French cache.
If the European region streaming media cache is not already delivering the requested stream, it connects to the streaming server in the U.S.A. region to obtain the stream and streams it to the cache that requested it.

Streaming misses sent to the U.S.A. region

The L4 switch in the U.S.A. region passes streaming misses from the European region directly to the streaming server. The L4 switch in the U.S.A. region passes requests from clients in the U.S.A. region to the L4 switch, and sends cache misses to the streaming server.

Failover

Failover works as follows for this scenario:

Current connections
If any of the streaming media caches goes down while it is delivering streams, all connections to clients that are viewing those streams are lost. The client must initiate the connection again. If the regional cache is still down when the client tries to initiate the connection again, the L4 switch passes the streaming request to the region above it in the hierarchy.
New requests
If the streaming media cache in a region goes down, the L4 switch passes any new requests to the region above it in the hierarchy.