Section E: Failover strategies

Do you need a failover strategy?

NetCache appliances are extremely reliable, with a proven track record of 99.9 percent availability. However, if you want to ensure uninterrupted content delivery service, Network Appliance recommends that you have a failover strategy to handle the unlikely occurrence of your GRM servers or center or edge caches failing.

Failover for GRM servers

To avoid interruption of your content delivery service, Network Appliance recommends that you deploy multiple GRM servers of the same type. Then, for example, if one GRM L7 server becomes unavailable, the other GRM L7 server can handle the traffic that would ordinarily be handled by the unavailable server.

All center and edge caches send status updates to all GRM servers listed in the cache's configuration. A GRM server that was temporarily unavailable begins receiving status information from the center and edge caches as soon as that GRM server becomes available again.

Failover from edge caches to center caches

GRM includes automatic failover to center caches if all edge caches in an edge group become unavailable at the same time. If multiple edge caches are deployed for an edge group, a client first fails over to the other edge caches returned by the GRM server as eligible caches. A client fails over to a center cache only if no edge caches are available.

If your CDN services both clients inside edge groups and clients not associated with edge groups, an automatic failover strategy will already be in place. The reason is that you will need to deploy at least one center cache to provide content delivery service for the clients outside of edge groups. That center cache automatically provides failover for the edge groups.

If your CDN does not service clients that are outside of edge groups, you should deploy a center cache specifically for failover if it is important for your CDN to provide uninterrupted service. A center cache can provide failover for multiple edge groups, possibly eliminating the need for deploying an extra edge cache in an edge group so that the traffic load can be handled even if one edge cache becomes unavailable.

If no center or edge caches are available

If no center or edge caches are available, the following occurs:

If the server is a GRM L7 server
The GRM L7 server can serve content to clients.
If the server is a GRM DNS server
By default, when no center or edge caches are available to serve the requested content, the GRM DNS server does not return any IP addresses to the local DNS nameserver. However, you can use the Origin Server Failover option in the GRM DNS server configuration to force the GRM server to query the back-end DNS nameserver and return the IP address of the origin server if a center or edge cache cannot be found to serve the requested content.