Using a Server Load Balancer for request distribution

When to use this method

If you are using any of the following client access methods, you can use a Server Load Balancer (SLB) to distribute requests over NetCache appliances:

Pointing clients to a single NetCache appliance
Using an automatic proxy configuration file
Using a nontransparent access method for access to a streaming media cache

Compare this discussion of using an SLB to Using DNS round robin for request distribution.

Methods SLBs use to distribute client traffic

SLBs are likely to provide strict load-balancing algorithms, for example:

Maximum connections
Weighted percentage
Fastest connection
Round robin

SLBs do not distribute requests based on IP address, which is desirable for caching service because distribution based on IP address results in a higher hit rate.

Example: using an SLB for Web caching

The following illustration shows a Web caching deployment that uses an SLB to balance Web requests over the NetCache appliances.

Although the previous illustration shows the SLB interacting with NetCache appliances functioning as Web caches, you can also deploy an SLB to balance requests over NetCache appliances configured as news caches and streaming media caches. The news caches would fetch cache misses from a news server rather than from the Internet.

To deploy an SLB with NetCache appliances, do the following:

Ensure that the SLB and NetCache appliances are on the same subnet.
On the SLB, specify the IP addresses of the NetCache appliances so that the SLB is aware of them.
Set up client browsers to point to the SLB rather than to the NetCache appliances. You can do this by having users configure their browsers to point to the SLB's IP address or, for Web caching, by entering the SLB's IP address in the automatic proxy configuration file.

Advantages of this method for request distribution

Using an SLB has the following advantages when compared to other methods:

Because the client browsers are unaware of the NetCache appliances, you can swap out NetCache appliances (for example, for maintenance) without affecting client requests.
If one NetCache appliance fails, the SLB shifts requests to the remaining NetCache appliances.
Requests are dynamically distributed over NetCache appliances

Disadvantages of this method for request distribution

Using an SLB has the following disadvantages when compared to other methods:

The cache hit rate is not as high as with methods that distribute requests based on IP address.
No failover to the Internet is available if all the NetCache appliances fail.