The Discoverability of the Web
- Anirban Dasgupta (Yahoo! Research)
- Arpita Ghosh (Yahoo! Research)
- Ravi Kumar (Yahoo! Research)
- Christopher Olston (Yahoo! Research)
- Sandeep Pandey (Yahoo! Research)
- Andrew Tomkins (Yahoo! Research)
Previous studies have highlighted the rapidity with which new content arrives on the web. We study the extent to which this new content can be efficiently discovered in the crawling model. Our study has two parts. First, we employ a maximum cover formulation to study the inherent difficulty of the problem in a setting in which we have perfect estimates of likely sources of links to new content. Second, we relax the requirement of perfect estimates into a more realistic setting in which algorithms must discover new content using historical statistics to estimate which pages are most likely to yield links to new content.
We measure the overhead of discovering new content, defined as the average number of fetches required to discover one new page. We show first that with perfect foreknowledge of where to explore for links to new content, it is possible to discover 50% of all new content with under 3% overhead, and 100% of new content with 28% overhead. But actual algorithms, which do not have access to perfect foreknowledge, face a more difficult task: 26% of new content is accessible only by recrawling a constant fraction of the entire web. Of the remaining 74%, 80% of this content may be discovered within one week at discovery cost equal to 1.3X the cost of gathering the new content, in a model with full monthly recrawls.