Among other things, an analysis was carried out in this project of "The functionality and effectiveness of Web caching servers", and "The requirements and recommendations of Web caching services". This has resulted in a report called "The Survey of caching requirements and specifications for prototype". Another report, "The evaluation of different caching strategies for national and international access", is in preparation.
For the daily analysis of the Web caching logfiles squid-stats v1.4 is used. This software is derived from the NLANR Cache Logfile Analysis Scripts.
An example of a Squid configuration file of an institute's Web caching server
is the one used by
Universiteit Utrecht. The one used by
Hogeschool van
Amsterdam is an example of a Web caching server which is configured as an
HTTP Accelerator.
At present, the Universiteit Utrecht Web Cache has two "siblings",
the HvA [www.hva.nl] and the SURFnet office [surver.surfnet.nl], and one
"parent", the National Top-level Web caching server
[cache.nic.surfnet.nl].
Both the caching servers of Universiteit Utrecht and of the SURFnet office cache all documents with the exception of documents in its own domain. The caching server of Hogeschool van Amsterdam, as it is configured as an HTTP Accelerator, also caches documents in its own domain, allowing it to be used by other Web caching servers (such as that of Universiteit Utrecht) as "parent" for the pages in the hva.nl domain.
With the exception of pages generated by CGI-scripts, each document retrieved by the National Top-level Web caching server is stored in its Cache, allowing other Web caching servers to use it as "parent" for pages both within and outside of the NL-domain. A mutual "parent" relation has been established with the Top-level Web caching server in Norway for the purpose of exchanging Web Documents.
![[Hierarchical Caching]](pictures/sn-proxy2.gif)
The present situation (April 17, 1997) of the SURFnet Web caching
mesh, with clickable items. In a hierachical model, a lower cache can
request documents from one of its "siblings", from its
"parent" or directly from the remote Web server.
If present, the local Web caching information pages are available behind the
clickable icons.
An example of an "Automatic Proxy Configuration" (APC) script is that developed by Universiteit Utrecht. The default extension of the file containing the script is "PAC". A special MIME content-type needs to be defined on the Web server containing this APC-script. For the Cern/W3C daemon the following should be added to the configuration file:
AddType .pac application/x-ns-proxy-autoconfig 7bitWhen the APC-facility is in use, failure of the first-named Web caching server in the APC-script will result in the Web browser automatically switching to the next in line. If the line is empty (none of the Web caching servers is available) the Web browser will request that the proxy servers, the Web caching servers here, be skipped and that the document be retrieved directly from the Web server.
An example of a user instruction Web page is that of the Universiteit Utrecht. It explains what a proxy server is, why it should be used, and which settings are needed for the Netscape Navigator.
For more information regarding Web caching and/or linking to the National Top-level Web caching server of SURFnet BV you can contact H.Bekker@cc.ruu.nl.
You will be sent a confirmation of registration, to which you are required to reply "OK" within 48 hours.
| Last modified: 17 april 1997 / Henny Bekker |
|