Determining the Need for Traffic Redirection

To optimize traffic, the appliance must intercept both the inbound and outbound packets for the flow.

Therefore, whenever you place an appliance out-of-path, you must redirect traffic from the client to the appliance.

There are three methods for redirecting outbound packets from the client to the appliance (known as LAN-side redirection, or outbound redirection):

•	PBR (Policy-Based Routing) — configured on the router. No other special configuration required on the appliance. This is also known as FBF (Filter-Based Forwarding).

If you want to deploy two Silver Peaks at the site, for redundancy or load balancing, then you also need to use VRRP (Virtual Router Redundancy Protocol).

•	WCCP (Web Cache Communication Protocol) — configured on both the router and the Silver Peak appliance. You can also use WCCP for redundancy and load balancing.

•	Host routing — the server/end station has a default or subnet-based static route that points to the Silver Peak appliance as its next hop. Host routing is the preferred method when a virtual appliance is using a single interface, mgmt0, for datapath traffic (also known as Server Mode).

To ensure end-to-end connectivity in case of appliance failure, consider using VRRP between the appliance and a router, or the appliance and another redundant Silver Peak.

How you plan to optimize traffic affects whether or not you also need inbound redirection from the WAN router (also known as WAN-side redirection):

•	If you enable subnet sharing (which relies on advertising local subnets between Silver Peak appliances) or route policies (which specify destination IP addresses), then you only need outbound redirection.

•	If, instead, you default to TCP-based or IP-based auto-optimization (which relies on initial handshaking outside a tunnel), then you must set up inbound and outbound redirection on the WAN router.

•	Additionally, for TCP flows to be optimized, both directions must travel through the same client and server appliances. If the TCP flows are asymmetric —as could occur in a high-availability deployment — you need to configure clusters for flow redirection among local appliances.

For more about flow redirection, see “Using Flow Redirection to Address TCP Asymmetry.”

A tunnel must exist before auto-optimization can proceed. There are three options for tunnel creation:

•	If you enable auto-tunnel on the Configuration - System page, then TCP-based or IP-based handshaking creates the tunnel. That requires outbound and inbound redirection to be in place.

•	You can let the Initial Configuration Wizard create the tunnel to the remote appliance.

•	You can create a tunnel manually on the Configuration - Tunnels page.

The following diagrams show where redirection is required and which methods you can use:

•	when subnet sharing is enabled

•	when using TCP-based or IP-based auto-optimization (that is, subnet sharing is not enabled)

•	when directed to a specific tunnel by the Route Policy

When using subnet sharing

n	Enable subnet sharing on both the local and remote appliances.

n	For outbound redirection to the out-of-path appliance (B), choose from PBR (or FBF), WCCP, or host routing.

n	Host routing only requires configuration on the client — not on the router or appliance.

 

When defaulting to TCP-based or IP-based auto-optimization

n	Initial handshaking between appliances happens outside the tunnel, requiring inbound redirection for packet routing.

n	For inbound and outbound redirection to the out-of-path appliance (B), choose from PBR (or FBF) or WCCP.

 

When specifying a tunnel

n	For outbound redirection to the out-of-path appliance (B), choose from PBR (or FBF), WCCP, or host routing.

n	With host routing, the outbound redirection is configured on the client, as opposed to on the router and/or appliance.

n	Host routing only requires configuration on the client — not on the router or appliance.

 

Please send comments or suggestions regarding user documentation to techpubs@silver-peak.com.