[SensorNetArch] Global Congestion Definition

[Joe Polastre]

I'm not sure that S-MAC or TRAMA are applicable for what we want to do.  The
idea is if you have some configurable set of parameters in SP, in other
words B-MAC plus some of the S-MAC primitives but in a configurable manner,
then FPS could use SP, perhaps transparently, to achieve efficient
end-to-end transfer.  S-MAC and TRAMA, as published, are (IMHO)
inappropriate for sensor nets due to their lack of control.  The B-MAC paper
pointed out that S-MAC could do significantly better by allowing a little
bit of control from higher layer protocols.  In turn, S-MAC and TRAMA as is
are not appropriate, but a version of them that interacts well with the
control and configuration ideas of B-MAC and SP may be appropriate.

Does that make sense?

-Joe

-----Original Message-----
From: sensornetarch-bounces at Millennium.Berkeley.EDU
[mailto:sensornetarch-bounces at Millennium.Berkeley.EDU] On Behalf Of Cheng
Tien Ee
Sent: Sunday, November 28, 2004 8:49 PM
To: David Culler
Cc: sensornetarch at Millennium.Berkeley.EDU
Subject: Re: [SensorNetArch] Global Congestion Definition

Hi,

Eric did point me to Barbara's FPS scheme, and I have a copy of her NSDI 
submission which I've looked at. I think the good points about the paper 
are that it considers flows at the network layer rather than just 
packets at the data-link layer, as well as fairness. But I believe that 
the duration of a cycle should not be a fixed network-wide parameter 
(it'll cause underutilization when traffic is low), it should be 
adaptive. I'm also not sure how FPS will perform if S-MAC is used, or 
maybe TRAMA.

Ee

David Culler wrote:

> I think you did a very nice job of seeding the discussion.  It would 
> make sense to look at Barbara Holte's Flexible Power Scheduling too.
>
> Cheng Tien Ee wrote:
>
>> Hi,
>>
>> I learnt a lot from today's meeting. There's this point I think I got 
>> wrong: the definition of global congestion. I would say that global 
>> congestion occurs when source motes generate packets over a period of 
>> time long enough for the congested mote to send feedback and 
>> rate-limit the sources. If the source motes don't generate traffic 
>> long enough for the feedback to reach them, then I would consider 
>> that to be local congestion. That's what I usually think when I hear 
>> the term.
>>
>> Sorry about the confusion.
>> Ee
>> _______________________________________________
>> SensorNetArch mailing list
>> SensorNetArch at Millennium.Berkeley.EDU
>> http://Mail.Millennium.Berkeley.EDU/mailman/listinfo/sensornetarch
>
>

_______________________________________________
SensorNetArch mailing list
SensorNetArch at Millennium.Berkeley.EDU
http://Mail.Millennium.Berkeley.EDU/mailman/listinfo/sensornetarch