Enabling Surrogate Computing

by
Sachin Goyal

Advised by
John Carter

Surrogate computing is the process of using the computing resources available in the environment as surrogates for the local device, i.e., to run tasks on your behalf. Potential uses include enabling pervasive computing (by offloading tasks from resource-constrained devices) and enabling novel wide area distributed services (by invoking computing jobs at selective locations on Internet).

We are building a surrogate computing platform that allows users to dynamically locate, allocate, and exploit resources on surrogate computers spread around the Internet. We are using virtual machine technology to enable the safe and effective sharing of surrogate resources while allowing clients to install and run arbitrary code on surrogates.

We will present the challenges we face in building a surrogate infrastructure and our ongoing ideas for how to tackle these problems. We will report the results of an initial experiment in which we use a surrogate to support real-time speech recognition on a handheld device (Sharp Zaurus), and present the response time and energy benefits of using our surrogate platform in that scenario.