In the event of power loss or failure, appliances enter bypass mode to avoid becoming a single point of failure in your network. If you want acceleration to continue, you can deploy redundant appliances as failover pairs.

In a physical in-path failover deployment, you configure a pair of appliances: one as a master, the other as a backup. The master, usually the appliance closest to the LAN, is active and the backup is passive. The backup becomes active if either the master fails or the master reaches its connection limit and enters admission control status. A backup doesn’t intercept traffic while the master appliance is active; it only pings the master to make sure it is online and processing data. If the master fails, the backup takes over and starts processing all of the connections. When the master comes back up, it sends a message to the backup that it has recovered, and the backup stops processing new connections but continues to serve old ones until they end.

In an out-of-path failover deployment, you deploy two server-side appliances and add a fixed-target rule to the client-side appliance defining master and backup target appliances. When both the master and backup appliances are functioning properly, the connections traverse the master appliance. If the master appliance fails, connections traverse the backup.

The master appliance uses an Out-of-Band (OOB) connection. The OOB connection is a single, unique TCP connection that communicates internal information only; it doesn’t contain optimized data. If the master becomes unavailable, it loses this OOB connection and the OOB connection times out in approximately 40 to 45 seconds. After the OOB connection times out, the client-side appliance declares the master unavailable and connects to the backup. During the 40 to 45 second delay before the client-side SteelHead declares a peer unavailable, it passes through any incoming new connections; they’re not blackholed.

While the client-side appliance is using the backup server-side appliance for optimization, it attempts to connect to the master appliance every 30 seconds. If the connection succeeds, the client-side appliance reconnects to the master for any new connections. Existing connections remain on the backup appliance for their duration. This is the only time (immediately after a recovery from a master failure) that connections are optimized by both the master and the backup appliances.

If both the master and backup appliances become unreachable, the client-side SteelHead tries to connect to both appliances every 30 seconds. Any new connections are passed through the network unoptimized.