33-3
Catalyst 3550 Multilayer Switch Software Configuration Guide
78-11194-09
Chapter 33 Configuring Web Cache Services By Using WCCP
Understanding WCCP
WCCP Message Exchange
This sequence of events describes the WCCP message exchange:
1.
The cache engines send their IP addresses to the WCCP-enabled switch by using WCCP, signaling
their presence through a Here I am message. The switch and cache engines communicate to each
other through a control channel based on UDP port 2048.
2.
The WCCP-enabled switch uses the cache engine IP information to create a cluster view (a list of
caches in the cluster). This view is sent through an I see you message to each cache engine in the
cluster, essentially making all the cache engines aware of each other. A stable view is established
after the membership of the cluster remains the same for a certain amount of time.
3.
When a stable view is established, the cache engine in the cluster with the lowest IP address is
elected as the designated cache engine.
WCCP Negotiation
In the exchange of WCCP protocol messages, the designated cache engine and the WCCP-enabled
switch negotiate these items:
•
Forwarding method (the method by which the switch forwards packets to the cache engine). The
switch rewrites the Layer 2 header by replacing the packet’s destination MAC address with the target
cache engine’s MAC address. It then forwards the packet to the cache engine. This forwarding
method requires the target cache engine to be directly connected to the switch at Layer 2.
•
Assignment method (the method by which packets are distributed among the cache engines in the
cluster). The switch uses some of the least-significant bits of the destination IP address to determine
which cache engine receives the redirected packet. The number of bits used is based on the number
of cache engines. If the number of cache engines is equal to a power of 2 (for example, 1, 2, 4 and
so forth), the switch evenly distributes (load balances) the traffic among the cache engines.
The switch does not support the mask assignment method described in the WCCP V2.0 Internet
Draft.
•
Packet-return method (the method by which packets are returned from the cache engine to the switch
for normal forwarding). These are the typical reasons why a cache engine rejects packets and
initiates the packet-return feature:
–
The cache engine is overloaded and has no room to service the packets.
–
The cache engine receives an error message (such as a protocol or authentication error) from the
web server and implements the dynamic client bypass feature. The bypass enables clients to
bypass the cache engines and to connect directly to the web server.
The cache engine returns a packet to the WCCP-enabled switch to forward to the web server as if
the cache engine is not present. The cache engine does not intercept the reconnection attempt. In this
way, the cache engine effectively cancels the redirection of a packet to the cache engine and creates
a bypass flow. The switch receives the returned packet through a generic-route encapsulation (GRE)
tunnel. The switch CPU uses Cisco express forwarding (CEF) to send these packets to the target web
server. When the server responds with the requested information, the switch uses the normal Layer
3 forwarding to return the information to the requesting client.