[net2-wg] Fwd: [Mantis-users] Current plans for MOS Release 1.1

[Philip Levis]

It sounds like the Mantis folks are developing some cool new  
protocols. Their announcement follows.

Phil

Begin forwarded message:

> From: Richard Han <rhan at cs.colorado.edu>
> Date: March 11, 2008 6:27:09 PM PDT
> To: mantis-users at mantis.cs.colorado.edu
> Subject: [Mantis-users] Current plans for MOS Release 1.1
>
> Just a quick update from our Mantis research team.  We're working  
> on improving Mantis OS for a release 1.1, perhaps sometime in May,  
> though more likely August.  We'll be planning to add the following  
> major features:
>
> (1) improved routing - a bidirectional version of the Cluster Tree  
> routing protocol (CTP), which we call CTP++.  It's already written,  
> and is undergoing testing.
>
> (2) a new transport layer (no name yet) - provides end-to-end ACKs  
> using a Stop-and-Wait protocol.  It's already written, and is being  
> tested on top of CTP++.
>
> (3) dynamic reprogramming at a thread granularity - we've adapted  
> Adam Dunkels' (of SICS) dynamic linking code to allow software  
> modules at the granularity of threads to be loaded at run time into  
> MOS.  Thus, you could replace on the fly an application thread that  
> is say sensing temperature with new code that provides more  
> accurate/calibrated sensing.  We thought this thread-level of  
> granularity in reprogramming ought to be sufficient in most cases.   
> The code is largely completed, but still has to be tested on top of  
> the transport protocol.  One hitch has been saving the new binary  
> image to flash, which brings us to...
>
> (4) a new flash file system - a simple file system for external  
> flash to allow storage of binary reprogrammed images, as well as  
> sensor data logs, and any other files that may be deemed  
> necessary.  The code is still being written.
>
> (5) a time synchronization protocol - a version of our TSync  
> protocol, designed for simple time synchronization.  We need it to  
> time stamp our sensor data, and also for accurate duty cycling, in  
> one of our deployments this summer.  It's written, but not yet  
> tested/integrated.
>
> (6) In addition, we are Deputizing MOS.  Deputy is a system from  
> Berkeley that adds memory safety checks in standard C code.  We've  
> adapted Deputy for MOS, and are annotating the code to add Deputy  
> checks for out-of-bounds memory writes.  Deputy checks at both  
> compile time and run time for bad memory writes.  We're thinking of  
> making Deputized code the default compile option for MOS, because  
> this enforces memory safety and is a good programming practice,  
> though we'd include a flag for compiling without Deputy.
>
> Some other capabilities that are probably beyond our current  
> available manpower, but highly desirable, include integrating  
> NodeMD into MOS to provide run-time checks for stack overflow and  
> deadlock/livelock during real in situ deployments.  This would be  
> extremely useful in deployments.  We haven't integrated X-MAC into  
> MOS yet.  This seems to be a useful duty-cycled MAC, and Adam  
> Dunkels tells us it's been integrated into Contiki already.  It  
> would be helpful to get a simulator like XMOS up and running, and  
> also allow TinyOS code to run in MOS using an updated version of  
> TinyMOS.
>
> Any feedback or volunteer help in testing or integrating new  
> features would be welcome.
>
> Rick
>
> -- 
> ----
> Prof. Rick Han
> University of Colorado,  Dept. of Computer Science
> (303) 492-0914    http://www.cs.colorado.edu/~rhan
> _______________________________________________
> Mantis-users mailing list
> Mantis-users at mantis.cs.colorado.edu
> http://mantis.cs.colorado.edu/mailman/listinfo/mantis-users