Troubleshooting guide

1-13
Cisco Wide Area Application Services Configuration Guide
OL-26579-01
Chapter 1 Configuring Traffic Interception
Configuring WCCP on WAEs
Destination IP address hashing guarantees that a single WAE caches a given file server. This method,
which allows a local coherency directive to be safely applied to the file server content (provided that no
other collaboration on the content occurs), improves performance and WAN link and disk utilization.
This method may distribute the load unevenly because of uneven activity on a file server.
Source IP address hashing has better potential for session distribution between the caches on
branch WAEs. This method may impact performance and WAN link and disk utilization (see the
previous description of factors to be aware of when load balancing is applied). Also, any change in the
IP address of a client (which can happen when working in DHCP environments) may cause the client to
switch to another branch WAE, which can cause the client to experience reduced performance until the
client’s working set is retrieved into the new cache.
Hashing that is based on a client IP address does not guarantee any locality of the hash key. For example,
clients from the same subnet (which are likely to share and collaborate on the same content) may be
assigned two different hash numbers and may be redirected to different branch WAEs, while clients from
different subnets may be assigned the same hash number and may be redirected to the same branch WAE.
Hashing that is based on a client IP address does guarantee consistency. For example, a client using the
same IP address is redirected to the same branch WAE.
In the service farm, a lead WAE is chosen to build the hash table that distributes the load between the
available WAEs. The lead WAE distributes the buckets evenly. The source IP address is hashed and the
resulting bucket determines the WAE that will handle the packet.
WCCP supports redirection by mask value assignments. This method relies on masking to make
redirection decisions. The decisions are made using special hardware support in the WCCP-enabled
router. This method can be very efficient because packets are switched by the hardware.
Note The masking method can only be used for load balancing with the Catalyst 3750, Catalyst 4500, and
Catalyst 6500 series switches, Cisco 7600 series routers, and Cisco ASR 1000 series routers. And, the
masking method can be used with the Cisco 2800, 3800, and 7200 series routers when they are running
Cisco IOS release 12.4(20)T or later releases.
You must explicitly specify masking. You can specify two mask values based on the source or destination
IP address of the packet. For WAAS, the default mask value is based on the source IP address. You can
enable masks by using the default values or specifying a particular mask. The default mask values,
specified in hexadecimal notation, are as follows:
• dst-ip-mask= 0x0
• src-ip-mask= 0xF00
You may specify the mask value with a maximum of seven bits. The WAE creates a table of the 2
7
(or
128) combinations, assigns the WAE IP addresses to them, and sends this table to the WCCP-enabled
routers. The router uses this table to distribute the traffic among all the WAEs that are in the service
group. Each packet that matches the WCCP service parameters is compared to this table and the packets
are sent to the matching WAE.
In a service farm where the WAEs have different masks, the first WAE to establish two-way
communication with the routers determines the farm’s mask. All other WAEs cannot join the farm unless
they are configured with the same mask.
Masking is typically used at the data center, where you can take advantage of the hardware accelerated
WCCP redirection capabilities of switches such as the Catalyst 6500 series switches. At the data center,
the load balancing goal should be to have all connections originating from a given client subnet
(typically equivalent to a branch) go to one data center WAE, to improve data redundancy elimination