In this work we evaluate the feasibility and performance of a low-power (a few Watts), wireless active gateway in servicing concurrent client requests of realistic web applications. Results show that the I/O-intensive Squid web caching can be well-supported on the low-power platform, with comparable performance and an order of magnitude better energy efficiency compared to Squid running on a conventional machine. On the other hand, we find that a computation-intensive service supporting enhanced human-computer interaction may lead to excessively long delays when hosted solely on our low-power active gateway. We further propose and evaluate an adaptive offloading technique to manage the computation load.
This work contributes to significant energy saving in the Internet-based information technology and consequently improves the sustainability of computing.