A Mobility Testbed
Enhancing Interactive Web Applications in Hybrid Networks
Aruna Balasubramanian Brian Neil Levine Arun Venkataramani appears at Mobicom 2008
Mobile Internet users have several options today including high
bandwidth cellular data services such as 3G, that may be the choice
for many. However, the ubiquity and low cost of WiFi suggests
an attractive alternative, namely, opportunistic use of open WiFi
access points (APs) or planned municipal mesh networks. Unfortu-
nately, for vehicular users, the intermittent nature of WiFi connec-
tivity makes it challenging to support popular interactive applica-
tions such as Web search and browsing. Our work is driven by two
questions. 1) How can we enable system support for interactive
web applications to tolerate disruptions in WiFi connectivity from
mobile nodes? 2) Can opportunistic mobile-to-mobile (m2m) trans-
fers enhance application performance over only using APs, and if
so, under what conditions and by how much?
We present Thedu, a system that enables access to Web search
from moving vehicles. The key idea is to use aggressive prefetching
to transform the interactive Web search application into a one-shot
request/response process. We deployed a prototype of Thedu on
the DieselNet testbed in Amherst, MA, consisting of transit buses
averaging 21 on the road at a time. Our deployment results show
that Thedu can deliver 4 times as many relevant web pages than not
using Thedu. A bus receives relevant web pages with a mean delay
of 2.3 minutes and within 0.55 minutes in areas with high AP den-
sity. Thedu augments AP connectivity with m2m transfers using
a utility-driven DTN routing algorithm and uses caching to exploit
query locality. Our analytic model and trace-driven simulations
suggest that m2m routing yields little benefit over using APs alone
even under moderately dense AP deployment such as in Amherst.
With sparsely deployed APs as may be the case in rural areas, our
conclusions are more mixed: m2m routing with caching improves
the number of relevant responses delivered per bus by up to 58%,
but the mean delay is significantly high at 6.7 minutes, calling into
question its practicality for interactive applications.