Net2 Meeting Notes for 1/29/2010
* CTP ETX representation * BLS
* Mike, JHU * Jorge, Berkeley * Phil, Stanford * Om, Stanford
* Jorge - URL with BLS code and information * Om - More tests with pre-tep123 and tep-123 CTP
CTP ETX representation
Om: CTP was putting EETX in the packets as link and path metric. This violates TEP123. ETX of 2 is represented by an EETX of 10. (subtract 1 and multiply by 10)
Om: I made two changes. Put ETX*10 in the packet. This is ETX representation. Also changed truncate in a few places in link quality calculation to round. For example, if the quality is 7.8, if we truncate (which is what happens when we perform integer division), we end up with 7 vs 8 when we round.
Om: I have committed the new changes to the CVS. With these changes, based on very preliminary results, the data delivery cost has gone down. This test was done on INDRIYA (sg).
Phil: Lets do the tests on Motelab and Tutornet.
Om: Yep, thats in my to-do list. Most of our understanding of CTP performance comes from Tutornet and Motelab so we need to run new experiments on those testbeds before concluding the performance with these changes.
Om: In the meantime, it will be great if people can take a look at the code and see if I missed anything when I made these changes in terms of representation and truncate/rounding issue.
Om: We have a change in plan for BLS:
Jorge: First status. The code has been ported to T2. Need to port to other platforms. Tutorial needs to be written up. The main issue is BVR is tightly integrated with BLS. I am guessing another 6 weeks of effort left. Then we need to test. Basically, lots of work left.
Jorge: I do not plan to continue on this project.
Om - Any relation with s4 at Sing?
Jorge - I grabbed a snapshot of that when I started.
Jorge - Main issue is bls+bvr data structure tight integration.
Jorge - I can create a tarball and post on my webpage.
Phil - Code with novel ideas and robust is great. Lots of people will use it. Large impact. But we can't release and support the code that is not ready.
Om - It will be useful to have a private URL that has a link to the source tarball, describes the status, provides some guidelines to whoever wants to work on it in the future, and also links to relevant papers. When someone asks me for a project, then I can suggest this project and send the URL.
Jorge - Will do.