Net2WG/Notes/20101112

From TinyOS Wiki
Jump to: navigation, search
    1. page was renamed from Net2WG/Notes/20101122

Contents

Net2 Meeting Notes 11/12/2010


Agenda


* A-MAC
* CTP bug reports

Participants


* Omprakash Gnawali, Stanford
* Stephen Dawson-Haggerty, UCB
* David Culler, UCB
* Philip Levis, Stanford
* Mike Jiang, JHU

Action Items


* Post bug reports on Google code (Om)
* A-MAC related discussion on Core (Phil)

Discussion Notes


A-MAC

Om: So, Stephen, tells us about HotMAC. You said you want to put in core at some point.

Stephen: I put it in contrib this week, I'm hoping we can... for stability, we were able to run CTP experiments for several hours without crashing. A-MAC/HotMAC is a receiver-initiated protocol I did for the CC2420 last November. It's much like RI-MAC: receivers beacon, and who wants to send responds to the beacon. Then there's a whole contention scheme and broadcast scheme. The big win is doing backcast, so a receiver knows whether anyone wants to transmit. The key thing is that it is really optimized for a bunch of corner cases.

Stephen: It works a lot better than LPL, especially under contention.

Om: So I guess that testing it with CTP is good, but I guess that this a core thing?

Phil: Yes.

Phil: So I think the right first step is to tell people it's in contrib, and try to get testers. Stephen, set up a web page. Then we can move it into core, as an option. After some more testing, maybe we can transition it to the main tree.

Om: What about interoperability?

Stephen: There is no RF230 version. I haven't tested on the micaz.

Om: OK, I can look into that once the web page is up.

Stephen: This thing totally bends the 15.4 standard in a lot of ways.

Mike: I can also put up the data from JHU. I can also run it on Motelab and post the results. I think we might have found one bug, I'll let you know what we find.

Om: What's the status of LPP?

Mike: I think LPP is in the tree right now.

Phil: No, it's in Razvan's tree. I think that it's OK to have multiple MAC protocols in the tree, but only if there are clear cases where each one is best.

Mike: From my perspective, AMAC uses some techniques from LPP. It might break some 15.4 rules, but it's like LPP.

Om: I'd also suggest one other argument -- if they are all robust pieces of software, we should keep them in the tree and make them available.

Mike: Also, I know LPP should work on more platforms than AMAC.

Phil: That's a good argument.

Om: Cool, so look forward to more discussions on this. I like this idea of having a clear statement of what are the defaults.

Phil: I'll put it on the core agenda.

CTP bug reports

Om: OK, let's talk about CTP bugs.

Phil: Why don't you forward me the emails.

Om: There are three. Number 1, link estimator keeps around dead nodes in the table. You are going to be sending estimates even after they are dead. The second is, if you use LPL and CTP, the ETX estimates for the path go way up. This is another bug that was reported. If you just have one hop, it's OK, but multihop it goes way up.

Phil: Is this because of hidden terminals?

Om: Maybe. The third one is that packets get stuck. This has been reported from 2 different experiments. The first is powernet, another person has reported a similar problem on tinyos-help. In that experiment, you have CTP and Drip, and after a while, CTP is not able to send packets. Drip does not stop sending packets.

Om: CTP doesn't seem to work when the network is dynamic.

Phil: Never supposed to. I think we should put these as Issues on tinyos-main.

Stephen: One thing, we might want a brief update on RPL next week.

Om: OK, that's the end of the meeting, we'll do some work offline.