Balancing computation and communication costs: The case for hybrid execution in sensor networks

Virtual machines (VM) are promising as system software in sensor networks. A major impediment to their widespread acceptance is their performance overhead. The compilation of VM byte code to native code improves performance, but increases the code’s footprint and cost of code distribution. Thus, there is an important tradeoff between the cost of computing and the cost of communication due to code distribution. We describe a remote Just-In-Time compilation service that is effective in combining interpretation with native execution to arrive at an efficient hybrid execution configuration. The principles may be applied to any middleware that is used to develop applications for sensor networks.