SOCIAL-BASED FORWARDING SCHEMES

Rance FredericksenCMPE 257

Wireless Networks

CONCEPT: SOCIAL BASED FORWARDING

Follows patterns of social interactions, generally taking after actual human interactions, to understand trends in mobility and communications

Forward based on social “popularity” of connected nodes—the more popular the recipient, the more likely the message will get do its destination while maintaining low overhead.

Forward based on community involvement—if the next hop is in the same social community as the destination, it is likely to be closer to the destination than other nodes.

Store, Forward, Store—if message is forwarded to another node in the community, a local copy is stored in the case the destination contacts the source.

SOCIAL METRICS

Group MembershipThis metric is simply a comparison of nodes and group affiliations. In human interaction, this is analogous to hobbies, clubs, etc, where like-minded people are likely to meet and socialize at length. This is implemented by having each node contain a social profile (Facebook?)

Centrality/PopularityBetweenness – This metric measures the extent to which a node lies on the shortest path linking other nodes (how often it lies “between” multiple nodes creating a shortest path. This can be measured dynamically by counting the number of unique contacts a node has over time intervals.

ALGORITHMSPROPOSED WORK

Social-Greedy ISocial-Greedy IISocial-Greedy III

Maximum Expected Social Welfare

LABEL

BUBBLE RAP

PERFORMANCE METRICS

Average Actually DelayThe average delay for all the delivered destinations to receive their data.

Delivery RatioThis is the ratio of number of nodes that have received deliveries to the total number of nodes. This metric is only valuable if sources and destinations are chosen randomly.

Average CostThe average number of relays (hops) used during the deliveries of all sources to all destinations.

Social-Greedy

I. If node u has a message for the destination v and encounters node w (who is “socially closer” to v than u), u hands off the message to w, but does not remove the local copy.

II. When node u (with message destined for v) encounters node w at time t0, u hands off message to w ( if w is “socially closer” to v than u). At any time after t0, u can only pass the message to an encountered node if the encountered node is “socially closer” to v than w.

III. Same as I, only removes the message from storage.

BUBBLE Rap

Foreach ( encounteredNode )If ( this.label == destination.label )

If ( encounteredNode.label == destination.label &&encounteredNode.localrank > this.localrank )encounteredNode.takeMessage()

elseIf ( encounteredNode.label == destination.label ||

encounteredNode.globalrank > this.globalrank )encounteredNode.takeMessage()

Assumptions: Each node is a member of at least one group

Each node contains a global and local ranking (local to their community/group)

Simulation in NS-2

Social-Greedy (I,II,III)

Node Mobility Social Profile Parameters (socialRank,socialNetwork)

Node Mobility Based on Social Profile Parameters (Not random)

BUBBLE RAP

Node MobilityVarying Communities/MembershipsRandom Node Social RankingsDynamically changing Communities/Memberships

Simulation in NS2Challenges

Social-Greedy (I,II,III)

Node Mobility Social Profile Parameters (socialRank,socialNetwork)

Node Mobility Based on Social Profile Parameters (Not random)

BUBBLE RAP

Node MobilityVarying Communities/MembershipsRandom Node Social RankingsDynamically changing Communities/Memberships

Simulation in NS-2

Position, Social Rank and Network Affiliation set randomly

Set up a routing agent in tcl/lib/ns-lib.tcl

Periodic beacons traverse the network to create/update route tables and social information (socialRank and socialNetwork)

For { set I } { I < $val(nn) } { incr I} { $ns at 0.00 "[$mnode_(0) set ragent_] sink"}

rt_lookup() function in SGRP Agent consults the route table and chooses the most applicable route to which to forward based on community/social data

ANTICIPATED ERROR

MobilityNodes are moving at random-the purpose of the social-based approach

is to assume non-random mobility, in terms of communities and grouping/mobility based on social likenesses

PopularityPopularity in this model is equal to the degree of the node. In the paper

on BUBBLE Rap, the authors suggest this is not the best good measure of popularity, and an average over a time interval is a better

approach of simulating a popularity metric

EXPECTED CONCLUSIONS

The Social-Greedy (I/II) algorithm should improve delivery ratio via the store, forward, store approach,

possibly showing dropped stored packets at the intermediate nodes (unless we presume infinite

storage)

In theory, the social-greedy (III) algorithm of forwarding messages based on belonging to the same

socialNetwork group will have no positive effect (because the group memberships were chosen at

random) unless I set a static map and group nodes by community organization. More to follow hopefully...

EXPECTED CONCLUSIONS

The BUBBLE Rap algorithm, that forwards based on between-ness averaged over time, however, will have a positive effect on delivery ratio, at the expense of

possible congestion at the “popular” nodes.

If node mobility can be tested (in the timeframe), I would like to see how these two algorithms stand up to each

other for delivery ratio and average cost

REFERENCES

Jahanbakhsh, Kazem. Shoja, Gholamali C. King, Valerie. Social-Greedy: A Socially-Based Greedy Routing Algorithm for Delay Tolerant Networks. MobiOpp '10 February 2010.

Lu, Mingming. Wu, Jie. Social Welfare Based Routing in Ad Hoc Networks. NSF grant ANI 0073736

Wei Gao, Qinghua Li. Bo, Zhao. Cao, Guohong. Multicasting in Delay Tolerant Networks: A Social Network Perspective. MobiHoc '09 May 2009.

Hui, Pan. Crowcroft, Jon. Eiko, Yoneki. BUBBLE Rap: Social-based Forwarding in Delay Tolerant Networks. MobiHoc '08 May 2008.