Modern computing models expected to accelerate growth of Sensor Networks

I just got back from San Francisco where I participated in a panel discussion at the ACM/IEEE International Conference on Information Processing in Sensor Networks.  On the panel with me were representatives from Sun SPOT (Arshan Poursohi), Sentilla (Cory Sharp), and Luminary Micro (Paul Kimelman).  The topic was: after 10 years of sensor network study, where we are and how do we move this forward.  Now, given the folks on the panel, you might have imagined a debate over the merits of .NET vs Java or Open Source vs proprietary platforms.  Instead, the panel was very uniform in their view that the technology will gain commercial traction to the degree that it is accessible from the higher level computing models and high productivity tools and we spent all of the very active discussion defending that view to the audience. 

This was obviously an academic audience.  They are used to working with very small platforms like TinyOS or proprietary platforms and the tools that go with them.  They work on 4 and 8 bit processors and in tiny amounts of memory.  In other words, they do the really hard work.  So, in fairness to them, I’m sure they didn’t want to hear that they were working harder than they need to and that they should just switch to one of the higher level platforms on a 32 bit processor.   But the uniform message from the panel was that for this technology to make the transition to the commercial space in any appreciable way, it needs to be much simpler.  And more than simpler, it needs to be familiar.  Paul made an interesting point.  You may have a high end light switch that does everything and runs on something like MF alongside a low end light switch that could run very well on TinyOS on an 8 bit processor.  But would you do that?  Would you have a separate development team using different tools or would you leverage the same logic that you used on the higher end device?  I am sure that there is some volume at which it is worth the duplication of resources but as prices for 32 bit processors come down, those volumes get larger and larger. 

One thing I think we did not get clear with the audience was how they could work with these higher level platforms.  Looking back on it, the discussion seemed to fall out as an ‘either/or’ dichotomy – “TinyOS vs the virtual machine”.   The virtual machines that provide the higher level of productivity and reliability are built on this lower level heavy lifting.  All of the thinking that was reflected in the talks that I saw in the afternoon before the panel is just as applicable.  There is still a need for securing the communication of nodes so that the message is non-reputable.  We still need small and efficient file systems.  We still need tools for diagnosing the network.  The point that I think we were all trying to make was that exposing those capabilities in the context of a high productivity environment like .NET will be necessary for commercial uptake. There are still lots of fundamental work to do to continue to improve the applicability of the higher level environments to the embedded space. At the same time, there is data that shows that two of the fastest growing languages in the embedded space are C#, and Java.

Comment Notification

If you would like to receive an email when updates are made to this post, please register here

Subscribe to this post's comments using RSS