Stigmergy Routing Protocol for Content-Centric Delay- Tolerant Networks (STIR)
Anh-Dung.Nguyen@isae.fr Patrick.Senac@isae.fr Michel.Diaz@laas.fr Contribution
We introduce STIR - a new routing protocol for Content- Centric Delay Tolerant Network. Nodes in the net- work, by contacts, form a gradient field like the stig- mergy of ants. This gradient field creates paths allow- ing data to efficiently reach destinations (in term of de- lay, hops, messages, buffering, compared to epidemic rout- ing), with the only knowledge of the node’s gradient value.
Reseach Context
The huge number of mobile wireless devices with their diversity and the explosion of the amount of information dramati- cally increased the complexity at the periphery of the Internet.
• Reducing the Internet’s core traffic while efficiently exploiting the Internet resouces at the periphery could be solved by deploying the Store and Forward communication between end-systems and objects → a form of ambient commu- nication autonomy
• The emerging Content Centric paradigm would make end the End-to-End paradigm. Users is interested only by the information content delivered by the network and not by its location.
Our contribution is at the crossing of
• Content Centric Network → reduce the complexity of information access
• Delay Tolerant Network → reduce the complexity of information communication ,→ Content Centric Delay Tolerant Network (CCDTN)
Question: How to define an efficient and simple routing protocol for CCDTN ?
Our solution
Our idea is to define a protocol which exploit the stigmergy between end-systems, build stochastic pheromone paths between the information and its users, then use these paths for information transmission. The protocol consist of 4 mechanisms:
a) Interest message diffusion:
Binary Spray And Wait algorithm
b) Gradient establishment:
Diffusion with constraint algorithm
b) Data content diffusion:
Binary Spray And Wait based algo- rithm
c) Gradient reinforcement:
Exponential Moving Average Smoothing technic
gnew = αgcontact + (1 − α)gold
Modeling
• On a square of size √
T × √
T , a set of N nodes ΦN = {n1, . . . , nN } i.i.d following a mobility model M
• A set of S (server) nodes ΦS ⊂ ΦN publish a ser- vice.
• A set of C (client) nodes ΦC ⊂ ΦN subcribe to this service
• The radio range is D and a encounter is realized when a node is in the radio range of another node
Assumptions
• The inter-contact time follows a exponential law of parameter
• At each contact, a node encounter only one another node
• The contact probabilies of differents nodes are equal(i.e. Pcontact(ni) = Pcontact(nj)∀(ni, nj) ∈ ΦN ).
Evaluation criterion
• Expected transmission delay
• Network capacity
• Buffering capacity
• Power consumption
Analytical analysis
Expected delay of Interest diffusion
E [DInterest(i)] = (N − 1)
i(N − i) + N − K − i
N − i E [DInterest(i + 1)] , i ∈
1, L 2
E [DInterest(i)] = (N − 1)
i(N − i) + N − K − i N − i
2i − L
i E [DInterest(i)] + L − i
i E [DInterest(i + 1)]
,
i ∈
L
2 + 1, L
E [DInterest(L)] = (N − 1) KL Expected delay of Data diffusion
E [DData(i)]g = (N − 1)
i(Mg − i) + Mg − i − 1
Mg − i E [DData] (i + 1)g
E [DData(Mg)] = (N − 1) Mg − 1
Simulation results
DTN simulator: TheOne
Parameters: number of nodes - 300 (1 interested nodes + 10 data carriers), mobility - RWP, density - ≈ 3000 nodes/Km2
Expected Delay Hop Count
,→ We got better results in case of using the gradient field !
On-going & Future works
• Study of mobility’s impact on routing performance
• Parametric mobility model respecting the locality principle
• Performance studies according to others criterion (e.g. buffer capacity, energy, . . .)
• Formal modeling of spatio-temporal metrics in dynamic networks
• Experimental testbed
References
[1] Van Jacobson et al Networking named content In CoNEXT 2009
[2] Chalermek Intanagonwiwat et al Directed diffusion for wireless sensor networking In IEEE/ACM Transactions on Networking 2003