Computer Science Department Christos Tsiaras tsiaras@aueb.gr Master Thesis Presentation (short edition) Efficient Minimization of Routing Cost in Delay Tolerant Networks Supervised by Dr. Stavros Toumpis toumpis@aueb.gr Zürich, 20 July 2011
Outline What are and how to model DTNs (focus on Deterministic DTNs) Why minimize routing cost in DTNs Our approach through a simple example Performance Potential uses Q & A Time Slide 1 of 26
What are DTNs Basic characteristics All nodes in DTNs can and should store data Efficient data forward Examples Military ad hoc networks Sensor networks Satellite networks Slide 2 of 26
How to model a DTNs Slide 3 of 26
Deterministic DTNs Full knowledge of network evolution Known topology @ every time slot Known connectivity state @ every time slot Known storage cost for every node @ every time slot Transmission cost for every possible link Slide 4 of 26
Why minimize routing cost in DTNs Nodes => Routing cost Time slots => Routing cost DTNs can be very patient consider forward data DTNs can be very impatient consider routing decisions Slide 5 of 26
Our approach through a simple example Slide 6 of 26
Our approach through a simple example Slide 7 of 26
Our approach through a simple example Slide 8 of 26
Our approach through a simple example Slide 9 of 26
Our approach through a simple example Slide 10 of 26
Our approach through a simple example Slide 11 of 26
Our approach through a simple example Slide 12 of 26
Our approach through a simple example Slide 13 of 26
Our approach through a simple example Slide 14 of 26
Performance Simple Dijkstra implementation O (T 2 V 3 net ) O (T 2 V 2 net ) many to many one to many Recursive use of Dijkstra O (T V 3 net ) many to many O (T V 2 net ) one to many Simple Bellman-Ford O (T 2 V 4 net ) O (T 2 V 3 net ) many to many one to many Recursive Bellman-Ford O (T V 4 net ) O (T V 3 net ) many to many one to many Slide 15 of 26
Performance Slide 16 of 26
Performance Slide 17 of 26
Performance Slide 18 of 26
Performance Slide 19 of 26
Performance Slide 20 of 26
Potential uses Distributed scenario Recursive Dijkstra Airplanes network! Shortest paths precalculation about black box data exchange between planes Most of the Black box data reach the ground even if plane does not! Slide 21 of 26
Potential uses Distributed scenario Recursive Bellman-Ford Distributed data replication Replicate data with minimum cost Maybe someone want to pay to transfer the trafic through his network (information?) Slide 22 of 26
Potential uses Centralised scenario Recursive Dijkstra WSN Plant smart nodes and create a backbone for the WSN e.g. Runners Protocol Slide 23 of 26
Demonstration of our implementation This is the fun part Slide 24 of 26
Demonstration of our implementation Question: Find the minimum cost journey, so node 1 will send data to node 10 at a 21 maximum number of time slots Answer: Node 1 will store data for 8 time slots where he will forward the data to node 8 Node 8 will store data for 4 time slots where he will meet node 10 and he will deliver the data Data will reach node 10 in 12 time slots All nodes will continue travelling untill time slot 21 Slide 25 of 26
Conclusions Every DTN should be handled with different routing algorithm Distributed and Centralised routing decisions should be possible Fast calculation of the optimal routing path is mandatory Slide 26 of 26
Q & A Time This presentation can be found at http://dl.dropbox.com/u/1507044/projects/tsiaras-zurich_2011.pdf