Subscription-enhanced content delivery

  • Authors:
  • Mao Chen;Jaswinder Pal Singh;Andrea LaPaugh

  • Affiliations:
  • Department of Computer Science, Princeton University;Department of Computer Science, Princeton University;Department of Computer Science, Princeton University

  • Venue:
  • Web content caching and distribution
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

In existing content delivery systems user accesses are popularly used for predicting the request pattern of contents. In novel web applications such as publish/subscribe services, users explicitly provide statements of interest in the form of subscriptions. These subscriptions provide another source of user information in addition to access patterns. This paper addresses the content delivery problem when user-stated interest is available. Each request by a user is either based on a notification about the availability of content that matches the user's subscriptions, or general browsing that is not based on the publish/subscribe service. We propose two approaches to content delivery that exploit both proactive push-time placement and passive access-time replacement based on the subscription information, the access pattern of subscribers, and that of non-subscribers. In our simulation-based evaluation, the two approaches are compared to an access-based caching only algorithm and to three approaches that were proposed for pure notification-driven accesses in our earlier study [5]. The results demonstrate that incorporating subscription information judiciously can substantially reduce the response time, even when only a small portion of accesses is driven by notifications and the subscription information does not reflect subscribers' accesses perfectly. To our knowledge, this work is the first effort to investigate general content delivery and caching enhanced by using subscription information.