Congratulations to Arpit Gupta and Xinyu Xing, who recently had papers accepted at the 2014 Passive and Active Measurements Conference! Arpit’s paper studies connectivity and peering and what ISOC has been calling “tromboning” (paths on the continent that detour through LINX in London or AMS-IX in Amsterdam). Xinyu’s paper studies inconsistent Web search results using a tool we built called Bobble.
The abstracts of the accepted papers are below. The final versions of the papers will be posted here shortly, and the papers will be presented in March 2014 in Los Angeles.
Peering at the Internet’s Frontier: A First Look at ISP Interconnectivity in Africa
Arpit Gupta (Georgia Institute of Technology), Matt Calder (University of Southern California), Nick Feamster (Georgia Institute of Technology), Marshini Chetty (University of Maryland, College Park), Enrico Calandro (Research ICT Africa), Ethan Katz-Bassett (University of Southern California)
Abstract. In developing regions, the performance to commonly visited destinations is dominated by the network latency to these destinations, which is in turn affected by the connectivity from ISPs in these regions to the locations that host popular sites and content. We take a first look at ISP interconnectivity between various regions in Africa and discover many circuitous Internet paths that should remain local often detour through Europe. We investigate the causes of circuitous Internet paths and evaluate the benefits of increased peering and better cache proxy placement for reducing latency to popular Internet sites.
Exposing Inconsistent Web Search Results with Bobble
Xinyu Xing (Georgia Institute of Technology), Wei Meng (Georgia Institute of Technology), Dan Doozan (Georgia Institute of Technology), Nick Feamster (Georgia Institute of Technology), Wenke Lee (Georgia Institute of Technology), Alex Snoeren (UC San Diego)
Abstract. Personalized Web search can potentially provide users with search results that are tailored to their geography, the device from which they are searching, and a variety of other preferences and predispositions. Although most major search engines employ some type of personalization, the algorithms used to implement this personalization remain a “black box” to users, who are not aware of the effects of these personalization algorithms on the results that they ultimately see. Indeed, many users may be unaware that such personalization is taking place at all. This papers take a first look at the nature of inconsistent search results that result from location-based personalization and search history. We present the design and implementation of Bobble, a tool that executes a single user query from a variety of different vantage points and under a range of different conditions and compared the consistency of the results that are returned from each query. Using more than 75,000 search queries from about 175 users over a nine-month period, we explore the nature of inconsistencies that arise in different search terms and regions and find that 98\% of all Google search queries from Bobble users resulted in some inconsistency, and that geography is more important than search history in influencing the nature of the inconsistency. Different from a recent study, our measurement also indicates that the influence of search history on search inconsistency is medium but not moderate. To demostrate the potential negative impact of search personalization, we also use Bobble to investigate more than 4,000 locally disreputable businesses. We find that more than 40 of these businesses for whom the negative search results are hidden from the local Google search result set but not in other Google search result sets obtained from other regions.