The simplest configuration is using a single cache server attached to a single news server and having several clients attached to the news cache. Other cooperation patterns among the caches are hierarchical news caches or a cooperating pool of caches. These organization types are called cascading and neighboring , respectively. Both patterns are already in use on the World-Wide Web. For a detailed discussion see [BDH+94]. We will discuss the usability and implications of these cooperation patterns in the following sections.
Cascading means to build a tree of news caches as shown in Figure 4.3. If a news cache at a certain layer does not provide the requested data, it forwards the request to its upstream news cache and so on. Since the news cache acts as an NNRP server and uses NNRP commands on the server side setting up a tree of news caches is easy.
A high tree usually brings only minor benefits, because the higher level caches may have to store so many articles that they may be replaced by ordinary news servers. Also the performance might degrade, because a request for an article will possibly have to be processed by each news cache possibly up to the root of the tree. Under these circumstances it is often better to set up a news server instead of one of the higher level caches and attach the other caches to this news server.
Neighboring is a technique where the cache server checks whether the requested document can be found in other caches before asking the information server. If one of these neighboring caches stores the data, the data will be retrieved from that cache server.
We think that it is not necessary to consider this structure for a News Cache currently. The infrastructure provided by the currently existing news servers is very dense and it is faster to query a news server than the neighboring caches.
