WCCP uses Cisco Content Engines (or other content engines running WCCP) to localize web traffic patterns in the network, enabling content requests to be fulfilled locally. Traffic localization reduces transmission costs and download time.

WCCP enables Cisco IOS routing platforms to transparently redirect content requests. The main benefit of transparent redirection is that users do not need to configure their browsers to use a web proxy. Instead, they can use the target URL to request content, and have their requests automatically redirected to a content engine. The word "transparent" in this case means that the end user does not know that a requested file (such as a web page) came from the content engine instead of from the originally specified server.

When a content engine receives a request, it attempts to service it from its own local cache. If the requested information is not present, the content engine issues its own request to the originally targeted server to get the required information. When the content engine retrieves the requested information, it forwards it to the requesting client and caches it to fulfill future requests, thus maximizing download performance and substantially reducing transmission costs.

WCCP enables a series of content engines, called a content engine cluster, to provide content to a router or multiple routers. Network administrators can easily scale their content engines to manage heavy traffic loads through these clustering capabilities. Cisco clustering technology enables each cluster member to work in parallel, resulting in linear scalability. Clustering content engines greatly improves the scalability, redundancy, and availability of your caching solution. You can cluster up to 32 content engines to scale to your desired capacity.

WCCP uses either generic routing encapsulation (GRE) or Layer 2 (L2) to redirect or return IP traffic. When WCCP forwards traffic via GRE, the redirected packets are encapsulated within a GRE header. The packets also have a WCCP redirect header. When WCCP forwards traffic using L2, the original MAC header of the IP packet is overwritten and replaced with the MAC header for the WCCP client.

Using L2 as a forwarding method allows direct forwarding to the content engine without further lookup. Layer 2 redirection requires that the router and content engines are directly connected, that is, on the same IP subnetwork.

When WCCP returns traffic via GRE, the returned packets are encapsulated within a GRE header. The destination IP address is the address of the router and the source address is the address of the WCCP client. When WCCP returns traffic via L2, the original IP packet is returned without any added header information. The router to which the packet is returned will recognize the source of the packet and prevent redirection.

The WCCP redirection method does not have to match the return method.

L2 forwarding, return, or redirection are typically used for hardware accelerated platforms. In Cisco IOS Release 12.4(20)T and later releases, L2 forwarding, return, and redirection can also be used for software switching platforms.

On Cisco Catalyst 6500 Switches with a PFC4, GRE decapsulation is supported in hardware.

On Cisco ASR 1000 Series Aggregation Services Routers, both the GRE and L2 forward/return methods use the hardware, so there is not any significant performance degradation between them.

For content engines running Application and Content Networking System (ACNS) software, use the wccp custom-web-cache command with the l2-redirect keyword to configure L2 redirection. For content engines running Cisco Wide Area Application Services (WAAS) software, use the wccp tcp-promiscuous command with the l2-redirect keyword to configure L2 redirection.

For more information on Cisco ACNS commands used to configure Cisco Content Engines, see the Cisco ACNS Software Command Reference, Release 5.5.13.

For more information on WAAS commands used to configure Cisco Content Engines, see the Cisco Wide Area Application Services Command Reference (Software Versions 4.2.1).

The WCCP Mask Assignment feature enables mask assignment as the load-balancing method (instead of the default hash assignment method) for a WCCP service.

For content engines running Application and Content Networking System (ACNS) software, use the wccp custom-web-cache command with the mask-assign keyword to configure mask assignment. For content engines running Cisco Wide Area Application Services (WAAS) software, use the wccp tcp-promiscuous command with the mask-assign keyword to configure mask assignment.

Cisco Catalyst 6500 series switches with a PFC4 support WCCP Mask assignment in hardware.

For more information on Cisco ACNS commands used to configure Cisco Content Engines, see the Cisco ACNS Software Command Reference, Release 5.5.13.

For more information on WAAS commands used to configure Cisco Content Engines, see the Cisco Wide Area Application Services Command Reference (Software Versions 4.2.1).

Catalyst 6500 series switches and Cisco 7600 series routers provide WCCP Layer 2 Policy Feature Card (PFC) redirection hardware acceleration. Hardware acceleration allows Cisco Content Engines to perform a L2 MAC address rewrite redirection method when directly connected to a compatible switch or router.

Redirection processing is accelerated in the switching or routing hardware, which is more efficient than L3 redirection with Generic Routing Encapsulation (GRE). L2 redirection takes place on the switch or router, and is not visible to the Multilayer Switch Feature Card (MSFC). The WCCP L2 PFC redirection feature requires no configuration on the MSFC. The show ip wccp {service-number | web-cache} detail command displays which redirection method is in use for each content engine.

In order for the router or switch to make complete use of hardware redirection, the content engine must be configured with L2 redirection and mask assignment.

Use the ip wccp web-cache accelerated command on hardware-based platforms to enforce the use of L2 redirection and mask assignment. Using this command configures the router to form a service group and redirect packets with an appliance only if the appliance is configured for L2 and mask assignment.

The following guidelines apply to WCCP Layer 2 PFC redirection:

• The WCCP Layer 2 PFC redirection feature sets the IP flow mask to full-flow mode.
• You can configure the Cisco Cache Engine software Release 2.2 or later releases to use the WCCP Layer 2 PFC redirection feature.
• L2 redirection takes place on the PFC and is not visible to the MSFC. The show ip wccp {service-number | web-cache} detail command on the MSFC displays statistics for only the first packet of an L2 redirected flow, which provides an indication of how many flows, rather than packets, are using L2 redirection. You can view information about L2 redirected flows by entering the show platform flow ip command. The PFC3 provides hardware acceleration for GRE. If you use WCCP Layer 3 redirection with GRE, there is hardware support for encapsulation, but the PFC3 does not provide hardware support for decapsulation of WCCP GRE traffic.

With WCCPv1, only a single router services a cluster. In this scenario, this router is the device that performs all the IP packet redirection. The figure below illustrates the WCCPv1 configuration.

Figure 1

WCCPv1 Configuration

Content is not duplicated on the content engines. The benefit of using multiple content engines is that you can scale a caching solution by clustering multiple physical content engines to appear as one logical cache.

The following sequence of events details how WCCPv1 configuration works:

1. Each content engine is configured by the system administrator with the IP address of the control router. Up to 32 content engines can connect to a single control router.
2. The content engines send their IP addresses to the control router using WCCP, indicating their presence. Routers and content engines communicate to each other via a control channel; this channel is based on UDP port 2048.
3. This information is used by the control router to create a cluster view (a list of caches in the cluster). This view is sent to each content engine in the cluster, essentially making all the content 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.
4. When a stable view has been established, one content engine is elected as the lead content engine. (The lead is defined as the content engine seen by all the content engines in the cluster with the lowest IP address). This lead content engine uses WCCP to indicate to the control router how IP packet redirection should be performed. Specifically, the lead content engine designates how redirected traffic should be distributed across the content engines in the cluster.

Multiple routers can use WCCPv2 to service a content engine cluster. This configuration is in contrast to WCCPv1, in which only one router could redirect content requests to a cluster. The figure below illustrates a sample configuration using multiple routers.

Figure 2

Cisco Content Engine Network Configuration Using WCCPv2

The subset of content engines within a cluster and routers connected to the cluster that are running the same service is known as a service group. Available services include TCP and UDP redirection.

Using WCCPv1, the content engines were configured with the address of the single router. WCCPv2 requires that each content engine be aware of all the routers in the service group. To specify the addresses of all the routers in a service group, you must choose one of the following methods:

• Unicast—A list of router addresses for each of the routers in the group is configured on each content engine. In this case the address of each router in the group must be explicitly specified for each content engine during configuration.
• Multicast—A single multicast address is configured on each content engine. In the multicast address method, the content engine sends a single-address notification that provides coverage for all routers in the service group. For example, a content engine could indicate that packets should be sent to a multicast address of 224.0.0.100, which would send a multicast packet to all routers in the service group configured for group listening using WCCP (see the ip wccp group-listen interface configuration command for details).

The multicast option is easier to configure because you need only specify a single address on each content engine. This option also allows you to add and remove routers from a service group dynamically, without needing to reconfigure the content engines with a different list of addresses each time.

The following sequence of events details how WCCPv2 configuration works:

1. Each content engine is configured with a list of routers.
2. Each content engine announces its presence and a list of all routers with which it has established communications. The routers reply with their view (list) of content engines in the group.
3. When the view is consistent across all content engines in the cluster, one content engine is designated as the lead and sets the policy that the routers need to deploy in redirecting packets.

WCCPv2 allows multiple routers to be attached to a cluster of cache engines. The use of multiple routers in a service group allows for redundancy, interface aggregation, and distribution of the redirection load. WCCPv2 supports up to 32 routers per service group. Each service group is established and maintained independently.

WCCPv2 provides optional authentication that enables you to control which routers and content engines become part of the service group using passwords and the HMAC MD5 standard. Shared-secret MD5 one-time authentication (set using the ip wccp [password [0 | 7] password] global configuration command) enables messages to be protected against interception, inspection, and replay.

If a content engine is unable to provide a requested object it has cached due to error or overload, the content engine will return the request to the router for onward transmission to the originally specified destination server. WCCPv2 provides a check on packets that determines which requests have been returned from the content engine unserviced. Using this information, the router can then forward the request to the originally targeted server (rather than attempting to resend the request to the content engine cluster). This process provides error handling transparency to clients.

Typical reasons why a content engine would reject packets and initiate the packet return feature include the following:

• Instances when the content engine is overloaded and has no room to service the packets
• Instances when the content engine is filtering for certain conditions that make caching packets counterproductive (for example, when IP authentication has been turned on)

The WCCP Mask Assignment feature enables mask assignment as the load-balancing method (instead of the default hash assignment method) for a WCCP service.

For content engines running Application and Content Networking System (ACNS) software, use the wccp custom-web-cache command with the mask-assign keyword to configure mask assignment. For content engines running Cisco Wide Area Application Services (WAAS) software, use the wccp tcp-promiscuous command with the mask-assign keyword to configure mask assignment.

Cisco Catalyst 6500 series switches with a PFC4 support WCCP Mask assignment in hardware.

For more information on Cisco ACNS commands used to configure Cisco Content Engines, see the Cisco ACNS Software Command Reference, Release 5.5.13.

For more information on WAAS commands used to configure Cisco Content Engines, see the Cisco Wide Area Application Services Command Reference (Software Versions 4.2.1).

The WCCP VRF Support feature enhances the existing WCCPv2 protocol by implementing support for virtual routing and forwarding (VRF).

The WCCP VRF Support feature allows service groups to be configured on a per VRF basis in addition to those defined globally.

Along with the service identifier, the VRF of WCCP protocol packets arriving at the router is used to associate cache-engines with a configured service group.

The interface on which redirection is applied, the interface which is connected to cache engine, and the interface on which the packet would have left if it had not been redirected must be in the same VRF.

In Cisco IOS Release 12.2(33)SRE, this feature is supported only on Cisco 7200 NPE-G2 and Cisco 7304-NPE-G100 routers.

This feature is supported only on Catalyst 6500 series switches with a PFC4.

In Cisco IOS releases that support the WCCP VRF Support feature, the use of GRE redirection results in the creation of new tunnel interfaces. You can display these tunnel interfaces by entering the show ip interface brief | include tunnel command:

Router# show ip interface brief | include tunnel
  
Tunnel0                172.16.0.1      YES unset  up                    up      
Tunnel1                172.16.0.1      YES unset  up                    up      
Tunnel2                172.16.0.1      YES unset  up                    up      
Tunnel3                172.16.0.1      YES unset  up                    up      
Router#

The tunnel interfaces are automatically created in order to process outgoing GRE-encapsulated traffic for WCCP. The tunnel interfaces appear when a content engine connects and requests GRE redirection. The tunnel interfaces are not created directly by WCCP, but are created indirectly via a tunnel application programming interface (API). WCCP does not have direct knowledge of the tunnel interfaces, but can redirect packets to them, resulting in the appropriate encapsulation being applied to the packets. After the appropriate encapsulation is applied, the packet is then sent to the content engine.

Note	The tunnel interfaces are not used to connect with incoming WCCP GRE return packets.

One tunnel is created for each service group that is using GRE redirection. One additional tunnel is created to provide an IP address that allows the other tunnel group interfaces to be unnumbered but still enabled for IPv4.

You can confirm the connection between the tunnels and WCCP by entering the show tunnel groups wccp command:

Router# show tunnel groups wccp

WCCP : service group 0 in "Default", ver v2, assgnmnt: hash-table 
   intf: Tunnel0, locally sourced
 WCCP : service group 317 in "Default", ver v2, assgnmnt: hash-table 
   intf: Tunnel3, locally sourced
 WCCP : service group 318 in "Default", ver v2, assgnmnt: hash-table 
   intf: Tunnel2, locally sourced

You can display additional information about each tunnel interface by entering the show tunnel interface interface-number command:

Router# show tunnel interface t0

Tunnel0
   Mode:multi-GRE/IP, Destination UNKNOWN, Source 10.1.1.80
   Application ID 2: WCCP : service group 0 in "Default", ver v2, assgnmnt: hash-table 
   Linestate - current up
   Internal linestate - current up, evaluated up

Router# show tunnel interface t1

Tunnel1
   Mode:multi-GRE/IP, Destination UNKNOWN, Source 172.16.0.1
   Application ID 2: unspecified
   Linestate - current up
   Internal linestate - current up, evaluated up

Router# show tunnel interface t2

Tunnel2
   Mode:multi-GRE/IP, Destination UNKNOWN, Source 10.1.1.80
   Application ID 2: WCCP : service group 318 in "Default", ver v2, assgnmnt: hash-table 
   Linestate - current up
   Internal linestate - current up, evaluated up

Router# show tunnel interface t3

Tunnel3
   Mode:multi-GRE/IP, Destination UNKNOWN, Source 10.1.1.80
   Application ID 2: WCCP : service group 317 in "Default", ver v2, assgnmnt: hash-table 
   Linestate - current up
   Internal linestate - current up, evaluated up
Router#

Note that the service group number shown in the examples is the internal tunnel representation of the WCCP service group number. Group 0 is the web-cache service. To determine the dynamic services, subtract 256 from the displayed service group number to convert to the WCCP service group number. For interfaces that are used for redirection, the source address shown is the WCCP router ID.

You can display information about the connected content engines and encapsulation, including software packet counters, by entering the show adjacency [tunnel-interface] [encapsulation] [detail] [internal] command:

Router# show adjacency t0  
           
Protocol Interface                 Address
IP       Tunnel0                   10.1.1.82(3)

Router# show adjacency t0 encapsulation 

Protocol Interface                 Address
IP       Tunnel0                   10.1.1.82(3)
  Encap length 28
  4500000000000000FF2F7D2B1E010150
  1E0101520000883E00000000
  Provider: TUNNEL
  Protocol header count in macstring: 3
    HDR 0: ipv4
       dst: static, 10.1.1.82
       src: static, 10.1.1.80
      prot: static, 47
       ttl: static, 255
        df: static, cleared
      per packet fields: tos ident tl chksm
    HDR 1: gre
      prot: static, 0x883E
      per packet fields: none
    HDR 2: wccpv2
       dyn: static, cleared
      sgID: static, 0
      per packet fields: alt altB priB

Router# show adjacency t0 detail 

Protocol Interface                 Address
IP       Tunnel0                   10.1.1.82(3)
                                   connectionid 1
                                   0 packets, 0 bytes
                                   epoch 0
                                   sourced in sev-epoch 1
                                   Encap length 28
                                   4500000000000000FF2F7D2B1E010150
                                   1E0101520000883E00000000
                                   Tun endpt
                                   Next chain element:
                                    IP adj out of Ethernet0/0, addr 10.1.1.82
Router# show adjacency t0 internal

Protocol Interface                 Address
IP       Tunnel0                   10.1.1.82(3)
                                   connectionid 1
                                   0 packets, 0 bytes
                                   epoch 0
                                   sourced in sev-epoch 1
                                   Encap length 28
                                   4500000000000000FF2F7D2B1E010150
                                   1E0101520000883E00000000
                                   Tun endpt
                                   Next chain element:
                                    IP adj out of Ethernet0/0, addr 10.1.1.82
                                    parent oce 0x4BC76A8
                                    frame originated locally (Null0)
                                   L3 mtu 17856
                                   Flags (0x2808C4)
                                   Fixup enabled (0x40000000)
                                         GRE WCCP redirection
                                   HWIDB/IDB pointers 0x55A13E0/0x35F5A80
                                   IP redirect disabled
                                   Switching vector: IPv4 midchain adj oce
                                   IP Tunnel stack to 10.1.1.82 in Default (0x0)
                                    nh tracking enabled: 10.1.1.82/32
                                    IP adj out of Ethernet0/0, addr 10.1.1.82
                                   Adjacency pointer 0x4BC74D8
                                   Next-hop 10.1.1.82
Router#

WCCP intercepts IP packets and redirects those packets to a destination other than the destination that is specified in the IP header. Typically the packets are redirected from a web server on the Internet to a web cache that is local to the destination.

Occasionally a web cache cannot manage the redirected packets appropriately and returns the packets unchanged to the originating router. These packets are called bypass packets and are returned to the originating router using either Layer 2 forwarding without encapsulation (L2) or encapsulated in generic routing encapsulation (GRE). The router decapsulates and forwards the packets normally. The VRF associated with the ingress interface (or the global table if there is no VRF associated) is used to route the packet to the destination.

GRE is a tunneling protocol developed by Cisco that encapsulates packet types from a variety of protocols inside IP tunnels, creating a virtual point-to-point link over an IP network.

In applications where packet flows are intercepted and redirected by a Cisco IOS router to external WCCP client devices, it may be necessary to block the packet flows for the application when a WCCP client device is not available. This blocking is achieved by configuring a WCCP closed service. When a WCCP service is configured as closed, WCCP discards packets that do not have a WCCP client registered to receive the redirected traffic.

By default, WCCP operates as an open service, wherein communication between clients and servers proceeds normally in the absence of an intermediary device.

The ip wccp service-list command can only be used for closed-mode services. Use the service-list keyword and service-access-list argument to register an application protocol type or port number.

When there is a mismatch between the service-list ACL and the definition received from a cache engine, the service is not allowed to start.

WCCP is a component of Cisco IOS software that redirects traffic with defined characteristics from its original destination to an alternative destination. The typical application of WCCP is to redirect traffic bound for a remote web server to a local web cache to improve response time and optimize network resource usage.

The nature of the selected traffic for redirection is defined by service groups specified on content engines and communicated to routers by using WCCP. The current implementation of WCCP in Cisco IOS releases prior to Cisco IOS Release 12.3(14)T allowed a maximum of eight service groups to be defined. This maximum restricted caching deployments. In Cisco IOS Release 12.3(14)T and later releases, the maximum number of service groups allowed across all VRFs is increased to 256.

WCCPv2 supports up to 32 routers per service group. Each service group is established and maintained independently.

WCCPv2 uses service groups based on logical redirection services, deployed for intercepting and redirecting traffic. The standard service is web cache, which intercepts TCP port 80 (HTTP) traffic and redirects that traffic to the content engines. This service is referred to as a well-known service, because the characteristics of the web cache service are known by both the router and content engines. A description of a well-known service is not required beyond a service identification. To specify the standard web cache service, use the ip wccp command with the web-cache keyword.

Note	More than one service can run on a router at the same time, and routers and content engines can be part of multiple service groups at the same time.

Figure 3

WCCP Service Groups

The dynamic services are defined by the content engines; the content engine instructs the router which protocol or ports to intercept, and how to distribute the traffic. The router itself does not have information on the characteristics of the dynamic service group’s traffic, because this information is provided by the first content engine to join the group. In a dynamic service, up to eight ports can be specified within a single protocol.

Cisco Content Engines, for example, use dynamic service 99 to specify a reverse-proxy service. However, other content engine devices may use this service number for some other service. The configuration information in this document describes how to enable general services on Cisco routers.

An interface may be configured with more than one WCCP service. When more than one WCCP service is configured on an interface, the precedence of a service depends on the relative priority of the service compared to the priority of the other configured services. Each WCCP service has a priority value as part of its definition. When an interface is configured with more than one WCCP service, the precedence of the packets is matched against service groups in priority order.

Note	The priority of a WCCP service group cannot be configured via Cisco IOS software.

With the ip wccp check services all command, WCCP can be configured to check all configured services for a match and perform redirection for those services if appropriate. The caches to which packets are redirected can be controlled by a redirect ACL as well as by the service priority.

If no WCCP services are configured with a redirect ACL, the services are considered in priority order until a service is found that matches the IP packet. If no services match the packet, the packet is not redirected. If a service matches the packet and the service has a redirect ACL configured, then the IP packet will be checked against the ACL. If the packet is rejected by the ACL, the packet will not be passed down to lower priority services unless the ip wccp check services all command is configured. When the ip wccp check services all command is configured, WCCP will continue to attempt to match the packet against any remaining lower priority services configured on the interface.

To redirect traffic using WCCP to a router running WAAS software that is also configured with NAT, enable the ip nat inside command on the WAAS interface. If you are not able to configure the ip nat inside command on the WAAS interface, disable Cisco Express Forwarding. You must also update the WCCP redirect ACL to include a private address to ensure that pretranslated traffic is redirected.

CPU usage may be very high when WCCP is enabled. The WCCP counters enable a determination of the bypass traffic directly on the router and can indicate whether or not high CPU usage due to enablement of WCCP is the cause. In some situations, 10 percent bypass traffic may be normal; in other situations, it may be high. However, any figure above 25 percent should prompt a closer investigation of what is occurring in the web cache.

If the counters suggest that the level of bypass traffic is high, the next step is to examine the bypass counters in the content engine and determine why the content engine is choosing to bypass the traffic. You can log in to the content engine console and use the CLI to investigate further. The counters allow you to determine the percent of traffic being bypassed.

1.    enable

2.    configure terminal

3.   Enter one of the following commands:

4.    ip wccp check services all

5.    ip wccp [vrf vrf-name ] {web-cache | service-number}

6.    exit

1.    enable

2.    configure terminal

3.    ip multicast-routing [vrf vrf-name] [distributed]

4.    ip wccp [vrf vrf-name] {web-cache | service-number} group-address multicast-address

5.    interface type number

6.    ip pim {sparse-mode | sparse-dense-mode | dense-mode [proxy-register {list access-list | route-map map-name}]}

7.    ip wccp [vrf vrf-name] {web-cache | service-number} group-listen

1.    enable

2.    configure terminal

3.    access-list access-list-number remark remark

4.    access-list access-list-number permit {source [source-wildcard] | any} [log]

5.    access-list access-list-number remark remark

6.    access-list access-list-number deny {source [source-wildcard] | any} | [log]

7.    Repeat some combination of Steps 3 through 6 until you have specified the sources on which you want to base your access list.

8.    ip wccp [vrf vrf-name] web-cache group-list access-list

9.    ip wccp [vrf vrf-name] web-cache redirect-list access-list

1.    enable

2.    configure terminal

3.    interface type number

4.    ip nat inside

5.    ip wccp service-number redirect in

6.    exit

7.    interface type number

8.    ip nat outside

9.    ip wccp service-number redirect in

10.    exit

11.    interface type number

12.    ip nat inside

13.    ip wccp redirect exclude in

1.    enable

2.    show ip wccp [ vrf vrf-name] [service-number | web-cache] [detail | view]

3.    show ip interface

4.    more system:running-config