Strategies for client access to an accelerator

Ways to provide client access to an accelerator

Two ways to provide client access to an accelerator are as follows:

Set up transparent proxying on the NetCache appliance and on an L4 or L7 switch or a WCCP router
Configure DNS to support an accelerator

Using transparent proxying to direct traffic to an accelerator

When you are using an L4 or L7 switch or a WCCP router to provide access to an accelerator, you deploy the switch or router close to the accelerator. (In contrast, when an L4 or L7 switch or a WCCP router is deployed with a forward proxy, it is located close to the clients.) The following illustration shows the deployment of an L4 or L7 switch or a WCCP router and an accelerator.

The switch or router uses its hashing function to determine the accelerator to which to send the request. If both accelerators fail, the switch or router sends the request to the Web server or streaming server (as indicated by the dashed line in the illustration).

Optimizing request distribution for an accelerator

You can set a switch or WCCP router to optimize distribution for acceleration, as described in Request distribution with an L4 or L7 switch and Request distribution with a WCCP router.

An L7 switch and noncacheable objects

If you deploy an L7 switch, you can configure it to directly access the server it is accelerating for objects that cannot be cached, for example, CGI and private pages.

Allowing HTTP QuickTime encapsulation

To allow QuickTime HTTP encapsulation, you must redirect TCP port 7070. However, the RealNetworks legacy protocol PNA also used TCP port 7070 and it is not fully supported by NetCache. Therefore, if you configure an L4 or L7 switch or a WCCP router to redirect TCP port 7070 requests to a NetCache appliance, the appliance might also receive PNA traffic. NetCache proxies all PNA traffic to the origin server.

Using DNS to direct traffic to an accelerator

If you do not use transparent proxying to provide access to an accelerator, you can configure DNS to support client access to the accelerator. When you set up DNS, you configure the accelerator's IP addresses and IP address aliases in DNS so that the accelerator can receive requests for the servers it is accelerating. You could set up DNS in the following ways:

The accelerator handles all requests for the servers it is accelerating.
Many organizations want the accelerator to handle all the requests for the Web or streaming server. However, you might want to split requests between the accelerator and the Web server or streaming server in the following situations:
- If the link between the client and the Web server or streaming server cannot handle the load of requests to the server
- If neither the Web server or streaming server nor the accelerator can handle the full load
Requests are split between the accelerator and the servers it is accelerating.
This type of deployment provides failover because all requests are sent to the Web server or streaming server if the accelerator is unavailable.

Some organizations add multiple accelerators over time, further splitting the load on the Web server or streaming server. The accelerator essentially becomes another Web server or streaming server and shares the load.

See the Guide to Caching Protocols and Services for more details about how to set up DNS to support an accelerator.