Wayne Salamonsen,
R&D Engineer,
Internet Research & Development Unit
National University of Singapore


	The Internet Research and Development Unit is located within the 
National University of Singapore and is involved in a number of projects 
related to proxies and proxy caching. These include, but are not limited to: 
the development of a PICS Aware Proxy, and the performance study of proxy 
caches.


PICS

PICS (Platform for Internet Content Selection) is a technology, recently
developed by W3C, which allows for the classification of URLs through the
use of ratings schemes and PICS labels (http://w3.org/PICS).  The PICS
Aware Proxy project involves the development of a proxy which is capable
of fetching and comparing PICS labels for URLs to allow for content
filtering and selection mechanisms to be applied on a large scale.
Development will involve making modifications to sources of an existing
proxy system such as Squid.

The use of a proxy for PICS based filtering and control has a number of
advantages. Firstly, by directing outgoing URL requests through the proxy,
filtering and selection mechanisms can be implemented on a large scale.
This has potential usage in companies, K-12 schools, educational or
government institutions wishing to make use of the PICS technology. It also
provides a 
far better method than the currently used methods of black and white filter
lists which, if filtering criteria are altered, can become unwieldy and
difficult to maintain. 

Secondly, it enables and promotes the use of proxy caching and proxy
hierarchical caching of Web pages to improve performance.  As the number of
people on the Internet continues to increase at an astronomical rate, the use
and benefit of caching proxies for performance reasons must also increase to
handle the decreasing worldwide bandwidth.  Packaging PICS with the proxy
builds more functionality into the proxy thus promoting their further and
increased use.

Thirdly, to improve the performance of the PICS module, a label cache must be
constructed within the proxy to cache PICS labels in a similar fashion to that
in which web pages are currently cached. This is an essential part of a PICS
aware proxy as multiple 3rd party PICS label bureaus may have to be consulted
over the Internet before a suitable PICS label is located for the URL in
question.  This makes label fetching a potentially more time consuming process
than standard URL fetching.  The caching of PICS labels within the proxy will
largely remove this extra overhead allowing for continuing high proxy
performance levels. 


Caching performance

The goals of Web caching are to improve the overall response time for
users, and to reduce wide-area Internet bandwidth usage. Highly configured
proxy farms have been built by Singapore ISPs to cater to the current
user-base of 150,000, however statistics regarding the performance and
effectiveness of the proxy servers are not available.
 
IRDU is interested in the performance measurement, and scalability issues
of proxy caches. We have embarked on a project with a new local service
provider to benchmark different proxy servers in different topology
configurations. Subsequently, we would deploy the proxy cache for their
5000 trial users, and analyse the statistics produced by the operational
proxy cache.
 
The outcome of the project would help us to understand better the
performance benefits of hierarchical caching, the necessary tools
(WebStone, Harvest) for benchmarking proxies, and the load balancing
methods (DNS, ICP) for proxy caches. With the experience gained though
this project, we intend to offer proxy cache performance measurement and
tuning services to the other local ISPs. Hopefully this would improve the
performance of their current proxy caches, and encourage positive views of
the use of proxy servers in Singapore.