A network is asymmetric when a client request and its server response don’t use the same path through the network. This asymmetric network configuration is common for:
|
•
|
The following diagram shows a sample asymmetric network. In this example, each server appliance sees only one direction of the traffic flow.
 |
Flow redirection removes the asymmetry locally by merging the traffic of an asymmetric flow into a single appliance. An appliance that handles both directions of traffic for a flow can then optimize the flow properly. Specifically, this sets the stage for TCP acceleration and CIFS acceleration.
With flow redirection, the appliance that receives the first packet — that is, the TCP SYN packet — owns the flow and eventually receives all of that flow’s traffic. To be able to redirect, appliances are configured into clusters, whereby they communicate with each other and keep track of flows. Any given appliance can own multiple flows and redirect others, depending on whether or not the appliance received the initial TCP SYN packet.
The client request — in the form of an initiating SYN packet — may be received from the WAN side or the LAN side. This results in two possible scenarios for evading asymmetry:
The assumptions is that flow redirection happens across a LAN environment. Redirection across a WAN is not supported.
In this scenario, the WAN initiates the flow. All traffic returned from the server is redirected to the appliance that first received traffic from the WAN.
 |
The default behavior is that all traffic returned from the WAN is always returned to the appliance that first received the traffic from the LAN, regardless of the route policy at the remote appliance.
 |
