Re: [nwcrg] [dtn] Network Coding and SATCOM - call for review

[Lloyd Wood <[e-mail]>]

Likely relevant to this draft:


Can network coding bridge the digital divide in the Pacific?, Spiedel et al.,
International Symposium on Network Coding (NetCod), Sydney, June 2015.


http://arxiv.org/abs/1506.01048
http://dx.doi.org/10.1109/NETCOD.2015.7176795
https://internetnz.nz/sites/default/files/IR201404%20-%20University%20of%20Auckland%20-%20FINAL%20REPORT%20April%202015.pdf
https://conference.apnic.net/data/41/apricot-talk_1451964216_1456023873.pdf

My take from years of using a very large number of TCP accelerators across satellite
links is that:

- the performance benefits from the local file/stream/chunk/'object' rapid fetch
  from Internet side and cached delivery that they offer for the satellite link
  will far outweigh benefits of network coding across the path and its virtual
  'erasure channel' for responsiveness, but that's hard to see from a throughput
  graph.


- the over-satellite-link FEC feature offered is rarely that useful in practice,
  as channel coding makes it largely redundant. If the PEP accelerator FEC becomes

  useful, the channel coding needs to be adjusted for the SNR, and that's often

  a local ACM problem, e.g. heavy rain fade at Ka-band, which the O3b example used
  in the paper suffers from.

- there's a whole bunch of tuning of traffic for the offered satellite link
  capacity (QoS, prioritisation, shaping, etc) brought to the table by PEPs
  that network coding cannot consider. Sometimes this gets called 'Orchestration',
  though that means different things to different communities.


oh, draft typo: writting->writing. Haven't had the time to reread
and think about the whole draft revision yet.

L.
 
Lloyd Wood [e-mail] http://about.me/lloydwood



________________________________