1. Routing in Delay Tolerant Network Qing Ye EDIFY Group of Lehigh University

2. Outline Introduction Routing Issue and the Model DTN Routing Approaches  Gossiping-like  Shortest-path-like Summary and Our view

3. Introduction – Concept Delay Tolerant Networks: networks suffer from frequent, long duration partitioning and the lack of guaranteed end-to-end path Examples:  Low Earth Orbiting Satellites  Interplanetary networks by satellites, spaceships, and base stations  Special Military Networks  Groups of Handheld devices for search and rescue teams

4. Introduction – Characteristics Heterogeneity:  A hybrid of wired and wireless links  Bandwidth, delay and capacity of links may vary  Different types end-devices  Node may move or not Is DTN a Mobile Ad hoc Networks ?? Larger end-to-end delay No fixed stable end-to-end path Up and down of links may be predictable, periodic, or opportunistic

5. Routing Issue Conventional Routing:  Internet: distance vector and link state  Wireless Ad hoc networks: table-driven and source-initiated on- demand  Find and take advantage of the best current available end-to-end path between source and destination DTN Routing: what if no end-to-end path exist in the network  So-called message ferry becomes important  Buffer size in the intermediate nodes become a precious system resource  Reliability (eventual packet delivery in lifetime) become a research goal  Hop-by-hop scheme becomes dominant, rather than end-to-end methods

6. DTN Routing Model DTN  a multigraph G(E, N), multiple paths may exist according to different types of links Contact: point-to-point opportunity, (src, dst,uptime, downtime, delay, capacity) Message: (src, dst, creationtime, lifetime,size) Storage: finite buffer in intermediate nodes, infinite buffer in destinations Routing: hop-by-hop store-and-forward approach, how to select the next hop

7. DTN Routing Approaches – First Contact (FC) Basic idea: a message is forwarded along a contact chosen randomly among all current available contacts. Or the first available contact is selected Similar approach: Gossiping Routing Properties:  Simple: routing is based on a local view  Small control overhead  Not a loop-free approach  Unreliable: routing decision may be incorrect

8. DTN Routing Approaches – Minimum Expected Delay (MED) Assumptions:  the up and down schedules as well as the delays of overall contacts between the source and the destination are known  the overall message sizes are known  Next-hop delay = waiting delay + average propagation delay + average transmission delay Basic idea:  Dijkstra’s shortest path algorithm is applied  choose the path with the overall smallest next-hop delays Properties:  Achieve the minimum average delays  Loop-free  Superior contact happens later would be ignored  Congestion may happen since the route is fixed, no load balancing

9. DTN Routing Approaches – Earliest Delivery (ED) Assumptions: same as the MED approach Basic idea:  calculate the next-hop delay in a real-time manner  always pick up the current contact with smallest delays as the next hop Properties:  Loop-free  May achieve the minimum delay  Good for new contact  Not consider the queuing delay and the constraints of finite storage space in nodes

10. DTN Routing Approaches – Earliest Delivery with Local Queuing (EDLQ) Assumptions:  same as the MED  the local queue/buffer information is also known Basic idea:  next-hop delay = waiting delay + queuing delay + propagation delay + transmission delay  same as the ED approach Properties:  take the local buffer into account  can not avoid buffer overload in next few hops

11. DTN Routing Approaches – Earliest Delivery with All Queues (EDAQ) Assumptions:  same as the MED  the info of overall queue/buffer usage is known Basic idea:  same as EDLQ Properties:  is able to avoid the buffer overload problem  may not be the optimal route since the future traffic info is unknown

12. DTN Routing Approaches – Linear Programming What if we know everything of DTN? DTN Routing  “Quickest delivery Problem”  Minimum Cost Maximum Flow Problem  Linear Programming approach is good But  too strong assumptions  centralized appraoch

13. DTN Routing Approaches – Simulation Results dtnsim: a java based DTN simulator

14. DTN Routing Approaches – Simulation Results

15. DTN Routing Approaches

16. Summary Pros:  study the basic DTN scenarios  discuss the DTN routing issues and model  propose two type of DTN routing approaches: gossiping- like and shortest-path like  design a new DTN simulator Cons: the optimal routing have to be decided with strong assumptions, which may not be Cons:  the optimal routing have to be decided with strong assumptions, which are not true in the real world  centralized approaches  the problem of none end-to-end path in DTN is not studied  No QoS support