Utility and Quality-of-Service Models for 
the Dissemination of WWW Resources

Marios D. Dikaiakos

 

Athena Stassopoulou

Dept. of Computer Science
University of Cyprus
PO Box 537, 1678 Nicosia
CYPRUS
Email: mdd@ucy.ac.cy
Dept. of Computer Science
Intercollege
Nicosia, CYPRUS
Email: athena@intercol.edu

 

In this  paper we introduce  a mathematical formulation to  address the problem of  Web-content selection in the context  of emerging services that disseminate   WWW   resources.  In these services, a content-distributor selects and multicasts periodically Web-content  to a number of subscribed clients. Typically, content-distributors are satellite  operators  and   subscribers  are  operators  of  wide-area WWW-caching schemes.

The  basic premise  behind  these services  is  that the  disseminated content covers  adequately the interests of  subscribers, improves the hit-ratio of installed  Web-caches and, therefore, relieves overloaded terrestrial  TCP/IP  connections.   The  soundness  of  this  premise and  the overall feasibility of the  proposed approach, however, depend on a number of issues, such as:

These   issues    should   be   taken   into    consideration   by   a content-distributor in its selection of Web-content for dissemination, as well  as in  its negotiation with  potential clients  about the pricing of the service.

To  address  these  issues,  we  propose  a  theoretical  modeling  of Web-content dissemination that takes into account subscriber-interests (profiles).  In particular, we introduce two metrics that measure the similarity between the disseminated content and subscriber profiles: resemblance and coverage. Furthermore, we define two alternative charging schemes: Usage- and Subscription-based pricing.

We show that the similarity metrics introduced can be used to define formally the notions of Client Utility and Quality of Service (QoS). Consequently, we employ these metrics to model mathematically Client Utility  and QoS  for  the two pricing schemes in question. Based on  our modeling,  we  propose  a framework  within  which  the distributor  and   its  subscribers  can  define   and  negotiate  the Web-content-dissemination service.  This framework entails three dimensions:

We use this negotiation framework to compare theoretically the two alternative pricing schemes considered and prove that Subscription-based pricing is preferrable over Usage-based pricing for the content-distributor.

Finally, according to  our  modeling framework,  we  formalize  the selection of Web-content  for dissemination as  a combinatorial optimization  problem. We prove  that this  problem is  NP-complete and develop an approximation algorithm to  resolve it.  We implement the algorithm and run experiments, which corroborate the soundness  of the proposed negotiation framework.