Special Content Gateway deployment scenarios

Content Gateway can be deployed in proxy clusters with failover features that contribute to high availability. The proxy can also be deployed in a chain, either with other Content Gateway proxies or third-party proxies. This section describes some examples of these deployment scenarios.

Highly available web proxy

A highly available web proxy provides continuous, reliable system operation.

Proxy high availability may be accomplished via a proxy cluster that uses various failover contingencies. Such deployments may involve either an explicit or transparent proxy configuration, load balancing, virtual IP addresses, and a variety of switching options. This section summarizes some possibilities for highly available web proxy deployments.

Using proxy

As previously mentioned for the explicit proxy deployment, clients are specifically configured to send requests directly to a proxy. The configuration can be accomplished manually, or via a PAC file or a WPAD server.

An explicit proxy deployment for high availability can benefit from the use of virtual IP failover. IP addresses may be assigned dynamically in a proxy cluster, so that one proxy can assume traffic-handling capabilities when another proxy fails. Content Gateway maintains a pool of virtual IP addresses that it distributes across the nodes of a cluster. If Content Gateway detects a hard node failure (such as a power supply or CPU failure), it reassigns IP addresses of the failed node to the operational nodes.

Active/Standby

In the simple case of an active/standby configuration with 2 proxies, a single virtual IP address is assigned to the virtual IP address “pool.” The virtual IP address is assigned to one proxy, which handles the network traffic that is explicitly routed to it. A second proxy, the standby, assumes the virtual IP address and handles network traffic only if the first proxy fails.

This deployment assumes the proxy machines are clustered in the same subnet, and management clustering is configured (that is, both proxies have the same configuration). Below is an example.

Active/Active

In an active/active configuration with 2 proxies, more than one virtual IP address is assigned to the virtual IP address pool. At any point in time, one proxy handles the network traffic that is explicitly directed to it. This deployment is scalable for larger numbers of proxies.

Clients requesting the IP address of a proxy can be crudely distributed using round robin DNS. Round robin DNS is not a true load balancing solution, because there is no way to detect load and redistribute it to a less utilized proxy. Management clustering should be configured.

An increase in the number of proxy machines makes the use of a PAC file or WPAD for specifying client configuration instructions convenient. A PAC file may be modified to adjust for proxy overloads, in a form of load balancing, and to specify website requests that can bypass the proxy.

As with the active/standby configuration, an available proxy can assume a failed proxy’s load. Below is an example.

With load balancing

A load balancer is a network device that not only distributes specific client traffic to specific servers, but also periodically checks the status of a proxy to ensure it is operating properly and not overloaded. This monitoring activity is different from simple load distribution, which routes traffic but does not account for the actual traffic load on the proxy.

A load balancer can detect a proxy failure and automatically re-route that proxy’s traffic to another, available proxy. The load balancer also handles virtual IP address assignments. Below is an example.

Integrated Windows Authentication is supported with a load balancer. Contact Forcepoint Technical Support for more information.

Using transparent proxy

In a transparent proxy deployment for high availability, traffic forwarding may be accomplished using a Layer 4 switch or a WCCP v2-enabled router. Routers or switches can redirect traffic to the proxy, detect a failed proxy machine and redirect its traffic to other proxies, and perform load balancing.

Using a Layer 4 switch

In one simple form of transparent proxy, a hard-coded rule is used to write a proxy’s Media Access Control (MAC) address as the destination address in IP packets in order to forward traffic to that proxy. Traffic that does not include the specified proxy address for forwarding is passed directly to its destination. See below for an example.

As described for the explicit proxy, virtual IP addresses can be used in this scenario to enhance availability in case a proxy machine fails.

Using a WCCPv2-enabled router

WCCP is a service that is advertised to a properly configured router, allowing that router to automatically direct network traffic to a specific proxy. In this scenario, WCCP distributes client requests based on the proxy server’s IP address, routing traffic to the proxy most likely to contain the requested information.

In a proxy chain

Content Gateway can be deployed in a network that contains multiple proxy machines, including one or more third-party proxies. A proxy chain deployment can involve different scenarios, depending on where Content Gateway is located in relation to the client. The proxy that is closest to the client is called the downstream proxy. Other proxies are upstream.

Below is a simple example of proxy chaining. On the left, Content Gateway is the downstream proxy. On the right, Content Gateway is upstream.

See Chaining Content Gateway with other proxies for specific instructions on using Blue Coat™ ProxySG™ or Microsoft™ Forefront™ Threat Management Gateway as the downstream proxy.

Content Gateway is downstream

A simple deployment has Content Gateway as the downstream proxy, closest to the client. In this scenario, Content Gateway security features are well positioned for maximum protection and network performance.

In this scenario, use of Content Gateway authentication to validate client credentials is preferred. You must disable authentication on the third-party proxy.

CONFIG proxy.config.http.forward.proxy_auth_to_parent INT 1

You can then use a transparent identification agent (for example, Logon Agent) to facilitate client identification.

Content Gateway can additionally send the client IP address to the upstream third- party proxy using the X-Forwarded-For HTTP header. This is set in the Content Gateway manager (Configure > Protocols > HTTP > Privacy > Insert Headers section).

The X-Forwarded-For HTTP header is the de facto standard for identifying the originating IP address of a client connecting through an HTTP proxy. Some proxies do not utilize the X-Forwarded-For header.

For information about installing and deploying transparent identification agents, see Deploying transparent identification agents and Installing web protection components.

Content Gateway is upstream

When Content Gateway is the upstream proxy, the downstream third-party proxy can perform authentication and send client IP and username information in the HTTP request headers. Content Gateway authentication must be disabled.

In this scenario, caching must be disabled on the third-party proxy. Allowing the third- party proxy to cache web content effectively bypasses Content Gateway’s inspection capabilities for any website that was successfully accessed previously from the third- party proxy.

Content Gateway can be configured to read authentication from the following proxies in the downstream position:

Blue Coat ProxySG - 210 and later

Microsoft Forefront TMG - MBE and later

For detailed configuration instructions for Blue Coat ProxySG and Microsoft TMG server, see Chaining Content Gateway with other proxies.

Proxy cache hierarchy

Another form of proxy chain is a flexible proxy cache hierarchy, in which Internet requests not fulfilled in one proxy can be routed to other regional proxies, taking advantage of their contents and proximity. For example, a cache hierarchy can be created as a small set of caches for a company department or a group of company workers in a specific geographic area.

In a hierarchy of proxy servers, Content Gateway can act either as a parent or child cache, either to other Content Gateway systems or to other caching products. Having multiple parent caches in a cache hierarchy is an example of parent failover, in which a parent cache can take over if another parent has stopped communicating.

As mentioned earlier, the increasing prevalence of dynamic, user-generated web content reduces the need for Content Gateway caching capabilities.

See Hierarchical Caching in the Content Gateway Manager Help.

SSL chaining

Routing SSL traffic in a proxy chain involves the same parent proxy configuration settings used with other proxy-chained traffic. You identify the ports on which HTTPS requests should be decrypted and policy applied when SSL is enabled in the Protocols > HTTP > HTTPS Ports option in the Configure tab. Parent proxy rules established in parent.config for HTTPS traffic (destination port 443) determine the next proxy in the chain for that traffic.

Enable the Configure tab Content Routing > Hierarchies > HTTPS Requests Bypass Parent option to disable SSL traffic chaining when all other traffic is chained.

If you want to exclude SSL traffic from the parent proxy and tunnel the traffic directly to the origin server, enable the Tunnel Requests Bypass Parent option in the Configure tab Content Routing > Hierarchies. This option can be used for any tunneled traffic.