A Modular Sensornet Architecture: Past, Present, and Future Directions Arsalan Tavakoli , Prabal Dutta , Jaein Jeong , Sukun Kim , Jorge Ortiz , David Culler , Phillip Levis ¡ , and Scott Shenker Berkeley EECS Dept. Berkeley, California 94720 {arsalan, prabal, jaein, binetude, jortiz, culler}@cs.berkeley.edu shenker@icsi.berkeley.edu UC CS Dept. Stanford, California 94305 pal@cs.stanford.edu ¡ Stanford Introduction Wireless sensornets provide an unprecedented opportunity to gather huge volumes of data about the physical world around us. Sensornets, however, have severe resource constraints, in terms of power, memory, and bandwidth, which make gathering and processing the data an extremely challenging problem. The rst wave of sensornet programmers dealt with these constraints by building tightly-integrated and monolithic system stacks. While the resulting systems were far more energy, memory, and bandwidth ef cient than traditional systems, they had two unfortunate properties. First, they were extremely dif cult to program, as they had to confront myriad new networking and sensing challenges and doing so required very low-level control. Programming entire systems using low-level C code is a daunting task for experienced programmers, and all but impossible for the intended users of sensornets: general scientists. Second, the implementations provided few clearly de ned internal
/lp/association-for-computing-machinery/a-modular-sensornet-architecture-past-present-and-future-directions-qS6GmgyjGa