Challenging VMs on Battery-Powered Embedded Devices

Many consumer devices, such as portable game consoles or cell-phones, can be described as battery-powered wireless embedded devices. Many of these are not taking advantage of virtual ma-chines, certainly not for their core tasks, instead relying on C or close derivatives for implementing their behaviour. However, faced with software that becomes more and more dynamic and hardware that is increasingly heterogeneous, the move to virtual machines is as necessary as it is unavoidable. From our own experience in developing software for battery-powered embedded devices we de-scribe features that these upcoming virtual machines should pos-sess in order to win over the embedded crowd and be a viable com-petitor against C, not just the only viable option. 1. Opening Remark This paper will not advance the state-of-the-art in virtual machine research. Instead it sketches the domain of embedded systems de-velopment where virtual machines are sorely needed but do not seem to be catching on yet. In this paper we enumerate what we think are some reasons why this is the case, without handing any solutions at the time. We think this can be of interest to the VMIL workshop, given the fact that virtual machine research is looking into embedded devices (see for example the invited talk on the Maxine virtual machine). We feel, however, that these efforts are not yet ready to replace the entire embedded device development. Our position statement is that embedded software development, es-pecially for battery-powered mobile devices, will embrace virtual machines when these virtual machines are modularized, allow de-velopers to take advantage of hardware features, and offer intro-spective and optimization possibilities. 2