In this scenario, the Web or streaming site is very big and very busy. One server cannot handle the amount of traffic. One solution is to have multiple servers handle the site and perform DNS round robin or server load balancing to distribute requests over the site. The problem with this method is the need to keep the replicas synchronized with each other.
Another solution is to use a single Web server or streaming server, as applicable, to handle the site and use multiple accelerators to accelerate the server. You would use one of the methods described in Section C, Direct (nontransparent) client access methods to distribute the requests over the accelerators. For example, you could use a Server Load Balancer (SLB) or transparent proxying with an L4 or L7 switch or WCCP router. The SLB, switch, or router would distribute the requests over the accelerators.
Replication works correctly in this case because the accelerators pick up new content as the objects' Time-To-Live in the cache expires. Even in the case of Web pages whose Time-To-Live is a few minutes, if the pages are accessed frequently, considerable load can be taken off the server if many requests are received within that period of time.
Note
NetCache does not delete an object from the cache when its Time-To-Live expires. The next time a request for an object is received, NetCache checks the server to determine whether a newer version of the object exists. If the object did not change and the server can understand If-Modified-Since queries, the verification of the object's status is less expensive than obtaining the object from the server.
The following illustration shows multiple accelerators accelerating a single server (Web server or streaming server). In this scenario, an SLB is used to distribute requests over the accelerators. Alternatively, you could use an L4 or L7 switch or a WCCP router.
If you are using DNS round robin and NetCache appliances, you can use NetCache takeover pairs to enable one accelerator to take over proxy services for another if one accelerator in the pair fails. See Section B, Failover by using NetCache appliance takeover pairs.
Copyright © 1994-2002, Network Appliance, Inc. All rights
reserved.
Part No. 210-00372
File: accel6.htm
Last updated: 11/13/02 18:11:42