Hi Miguel,
A lot of the analysis I've done on MAC type stuff in the
past have been tools I built for the task at hand. I
agree, though, that analysis tools like this should be
developed and shared by the community as a general way to
make performance metrics. Unfortunately, I don't have the
time right now to make the tools I've built more general
purpose.
Are you running your software on actual nodes, or just a
simulator? My experience has been with only actual nodes,
and I haven't used a simulator before.
To test throughput, I made a program where one node would
continuously transmit messages back-to-back while another
node would continually receive messages. The payload of
the transmitter's messages would contain the number of
packets per second it was able to send in the last second.
The receiver would periodically (once every 10 seconds)
send out a message with the average number of packets per
second received over the last 10 seconds. With this, I
was able to determine if there were any low-level radio
send/receive discrepancies and find out how fast the
system is.
Energy testing is another big thing we're concerned with
here, and the only accurate way to do that is on actual
nodes. We have a 50kSps 24-bit DAQ with matlab to perform
detailed energy analysis. I find that a multi-meter isn't
nearly as accurate enough unless the node is stuck in a
particular state, but even then, it may be difficult to
get down into the uA resolution.
For fairness testing, we setup a few nodes that broadcast
to a base station as fast as possible. Each node keeps
track of how many times it has tried to broadcast, and
includes that number as part of the payload. After
leaving this running for some amount of time, we can see
if some nodes are getting dropped packets or pushed aside
more than others.
We also test range, which is more low-level radio testing
than MAC layer testing. That involves the receiver
keeping track of how many messages it has received and
comparing it to how far away and how high the node was off
the ground, as well as what type of surface it was above.
Hope this gives you some ideas, but unfortunately a lot of
these tools (from my knowledge) need to be written from
scratch. Someone should definitely make general purpose
evaluation tools available to the community if the tools
don't exist already. It will probably be a few months
before my team can produce them if we do.
-David
On Tue, 9 Jan 2007 16:56:23 +0000
"Miguel Pereira" <[EMAIL PROTECTED]> wrote:
Hi,
I´m on final tests of design a new MAC to Wireless
Sensor Networks. I steel
need to make some improvements and some test but what I
want for now is to
make some performance measure to identify the problems
and know where to
make the improvements. Although I don't know which
software could I use. I
would like to measure the parameters like energy
consumption, throughput,
latency, fairness, etc. I think that TOSSIM can't
measure all this points.
Even it can, I can´t use it because CC2420_lpl it is not
supported. I
already saw the ns-2 simulator but I need to construct
all the code again. I
have code only for TinyOS and what i need is something
like TOSSIM that use
the code of TinyOS and make the simulation. Can anyone
give me some help?
Hi, David Moss, can you explain to me how do you make
the performance
measures of your modules?
Thanks for your help and any elucidation
Miguel Pereira
_______________________________________________
Tinyos-help mailing list
[email protected]
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
