A Gossip Protocol for Subgroup Multicast

Jenkits, Hopkinson, Birman

gossip protocol multicast reliable data dissemination group

@inproceedings{jenkins2001gossip,
  title={A Gossip Protocol for Subgroup Multicast},
  author={Jenkins, K. and Hopkinson, K. and Birman, K.},
  booktitle={International Workshop on Applied Reliable Group Communication},
  year={2001},
  organization={{IEEE} Computer Society}
}

Apparent major approach to dealing with different groups of gossiping users is to just split the gossip networks entirely---don't deal with nodes that aren't interested in your data

• Makes sense in wired environment where you are unicasting to known nodes
• In a group of N machines, if one machine starts out with a novel piece of information, it takes O(logN)

Interested in trading off between quality/timeliness of updates, and gossip overhead

• E.g., if node is member of many groups and cannot handle load of all of them, wants to just monitor some

Modifies probability by which a node will infect another node

• Rather than standard uniform distribution
• Similar to biasing by network performance/connectivity?
• Modifies susceptibility based on requested update timeliness

Exchange of infectivity/susceptibility rates?