1 reputation
2
bio website stackoverflow.com/users/1140105/mappu
location New Zealand
age 24
visits member for 2 years, 10 months
seen Jan 30 '12 at 5:26

Software engineer, currently based in Christchurch, New Zealand.


Jan
29
comment Strategies for preventing isolated nodes in a dynamically changing undirected cyclic graph
I agree my worries are based around the discovery protocol.. I had envisaged a system where nodes will appear at random intervals, attempting to form edges with whatever nodes they recall from their previous session, and was trying to consider a discovery techniques that would ensure no nodes become isolated. I realise now that probably means there are too many variables to find a solution, so perhaps i will look into relevant papers and heuristic simulation of discovery protocols from here on. Thank you for helping me understand my own question! :(
Jan
28
comment Strategies for preventing isolated nodes in a dynamically changing undirected cyclic graph
Terribly sorry, thank you for helping me make this clear. Peer is vertex, Neighbour means another vertex one edge away. Peers disappear without warning (removing all their edges). My current approach to discovery and connectivity is to have peers start with a short pre-shared list of known peers, attempt to connect to them, (probably) find out they're full (Y neighbours already) but receive a list of their neighbours, and proceed to sort of spiral outward from the 'center' in this way (until they have at least X neighbours). Then, the known-peer list is cached between sessions.
Jan
28
awarded  Editor
Jan
28
revised Strategies for preventing isolated nodes in a dynamically changing undirected cyclic graph
clarified title + actual question
Jan
28
asked Strategies for preventing isolated nodes in a dynamically changing undirected cyclic graph
Jan
28
awarded  Autobiographer