Net2WG 20060921 Meeting Note
*Link Layer Abstraction *Collection Protocol
Arsalan (A), Phil (P), Rodrigo (R), Sukun (S), Note: Sukun
P: Hotnet paper about IPv6 over 802.15.4, reception rate for a link A->B, acknowledgement from A->B was higher than data from A->B. Data size is also a factor. But we don’t send a packet until we get a clear channel. If you look at bidirectional PRR, it is very pessimistic. Most accurate estimate of ETS is when you send a data packet. The general approach taken is link estimate is a moving average: EWMA. There are two places to obtain estimation: data traffic or beacon.
R: We talked about this in debugging collection protocol.
P: Exponentially moving average of beacon PRR.
(talks about how to estimate link, but I could not clearly follow)
R: If you have enough data packet, we can have enough beacon packets. However, in case of Trickle, beacon rate can be 1pkt/30min.
R: It would be worthy collecting data from multiple network configurations.
P: There is not much study about weight of moving average.
R: It would be quite straight forward to come up with some test.
R: What is the release target?
R: Is CTP in release?
P: Deluge is more urgent. Taking the current deluge code, port to T2.
R: But the network part could be different.
P: It might not.
R: There are a bunch of changes in network, but not very difficult. Is storage more difficult?
P: Yes, more significant change in storage. Prabal and Jonathan are working on storage. Once it is done, porting deluge might not be too difficult.
R: What is the status of CTP.
P: I have tested with 4 nodes, we get packets.
R: In terms of code, anything is missing?
P: Link estimator. I haven’t seen Congestion bit. How would it work? Kyle put a congestion bit, I put timeout(?). Om is putting link estimator part.
R: Are they on CVS?
P: We can start testing RSCTP. A former UCB undergraduate student, now a graduate student in UCSD is trying a better dissemination.
P: What is the topic next week?
A: I have TEP for link layer.