an and d Im Impl plementa ementation tion (01 0120 20442 - PowerPoint PPT Presentation

Net etwork work Ke Kerne nel l Ar Archi chitect tectures ures an and d Im Impl plementa ementation tion (01 0120 20442 4423) ) Ro Routing uting Chaiporn Jaikaeo chaiporn.j@ku.ac.th Department of Computer Engineering Kasetsart University Materials taken from lecture slides by Karl and Willig

Overv Ov rvie iew Uni nicast ast rout utin ing g in in MA MANETs NETs  Energy efficiency & unicast routing  Multi-/broadcast routing  Geographical routing  2

Un Unic icast, ast, ID ID-Cent ntric ric Ro Routing uting Given: a network/a graph   Each node has a unique identifier (ID) Goal: Send a packet from one node to  another  The routing & forwarding problem  Routing: ing: Construct a table telling how can reach a given destination  Fo Forw rwardi rding: ng: Consult this table to forward a given packet to its next hop Challenges  3

Cha hallen llenges ges in in WS WSNs Ns/MANETs MANETs Nodes may move around, neighborhood  relations change B C A Optimization metrics may be more  complicated  Not just “smallest hop count” 4

Ad ho Ad hoc c Ro Routing uting Pro rotocols tocols Because of challenges, standard routing  approaches not really applicable  Too big an overhead, too slow in reacting to changes  Examples: Dijkstra, Bellman-Ford Simple solution: Flo loodin ing   No routing table needed  Packets are usually delivered to destination  But: overhead is prohibitive  Usually not acceptable in most cases 5

Go Gossiping siping Needs no routing table   Similar to flooding Nodes forward packets with some  probability Haas et al. studies gossiping behavior and  found that  There is a critical probability, x  p < x : gossip dies out very quickly  p > x : gossip reaches most nodes 6

Ro Routing uting Pro rotocol tocol Cla lassif ification ication Main question: Wh When does the routing  protocol operate? Option 1: Alw lways ays tries to keep routing data  up-to-date  Protocol is proa oact ctive ive / tabl ble-drive driven Option 2: Route is only determined when  actu tually ally ne needed  Protocol operates on on demand nd Option 3: Combine these behaviors   Hybr brid protocols 7

Ro Routing uting Pro rotocol tocol Cla lassif ification ication Which data is used to identify nodes?   An arbitrary identifier?  The pos osition on of a node? Can be used to assist in ge geogr graphi hic c routing  protocols  Identifiers that are not arbitrary, but carry some structure? As in traditional routing  Structure akin to position, on a logical level?  8

Pro roactiv active e Pro rotocols tocols – Ex Exam ample ple Fisheye State Routing (FSR)   Basic observation: When destination is far away, details about path are not relevant  Look at the graph as if through a fisheye lens  Regions of different accuracy of routing information  LS information about closer nodes is exchanged more frequently 9

Re React active ive Pro rotocols tocols – Ex Exam ample ple Recall reactive routing protocols   Initially, no information about next hop is available at all  One possibility: Send packet to everyone one  Hope: At some point, packet will reach destination and an answer is sent pack – use this answer for ba back ckward learni ning ng the route from destination to source Examples   Ad Ad ho hoc O c On-dem emand and Distance nce Vect ctor or (AO AODV)  Dynamic amic Sou ource ce Rou outing (DSR) 10

DSR DS Dynamic Source Routing protocol  Use separate rout ute requ quest st/rou route te reply ly  packets to discover route  Data packets only sent once route has been established  Discovery packets smaller than data packets Store routing information in the discovery  packets 11

DS DSR R Ro Rout ute e Di Disco covery very Search for route from 1 to 5 [1] [1,7] 2 2 1 1 [1] 7 7 [1,7] 5 5 4 4 3 3 6 6 [1,4] 2 1 2 1 [1,7,2] 7 7 [1,4,6] 5 5 4 4 3 3 6 6 [5,3,7,1] [1,7,3] Node 5 uses route information recorded in RREQ to send back, via source routing , a route reply 12

AOD AODV V Ad hoc On-demand Distance Vector  Very popular routing protocol  Same basic idea as DSR for discovery  procedure Nodes maintain routing tables instead of  source routing 13

Al Alte ternative rnative - Ru Rumo mor r Ro Routing uting Think of an “agent” wandering through the  network, looking for data/events Agent initially  perform random walk Leave “traces”  in the network Later agents  can use these ? traces to find data 14

Overv Ov rvie iew Unicast routing in MANETs  En Energy gy effic icie ienc ncy y & uni unicast st rout uting ing  Multi-/broadcast routing  Geographical routing  15

Energy En rgy-Efficient Efficient Un Unic icast: ast: Go Goal als Minimize energy/bit  4 A 2  Eg., A-B-E-H 3 1 Maximize  1 2 C network B 3 2 "lifetime" D 1  Time until first 2 4 2 E F 3 node failure 2 G 1 2  loss of coverage 2 4  partitioning H Example: Send data from node A to node H 16

Ba Basic ic opt ptio ions ns fo for path r path me metr trics ics Max total available  battery capacity 4 A 2 Sum of batt. levels 3  without needless 1 detours 1 2 Example: A-C-F-H C  B 3 Min battery cost 2  D 1 Sum of reciprocal  battery levels 2 4 Example: A-D-H 2  E F 3 2 Min-Max batt. cost G  1 2 Largest reciprocal  2 4 level of all nodes in path H Minimize variance in  power levels 17

Ov Overv rvie iew Unicast routing in MANETs  Energy efficiency & unicast routing  Mul ulti ticast/ cast/broadcast broadcast rout utin ing  Geographical routing  18

Br Broadcas oadcast t & Mu & Mult lticas icast Distribute a packet to all reachable nodes  ( br broadcas dcast ) or to a subgroup ( mul ulticast icast ) Basic options   Source-based tree: one tree per source Minimize total cost  Minimize maximum cost to each destination   Shared, core-based trees  Mesh Provides redundancy in data transfer  19

Go Goals als fo for So r Sourc urce-Based Based Tr Trees For each source,  Steiner tree minimize total al cost st Src Dest 2 2  The Steiner tree problem 2 1 For each source,  minimize max axim imum um Dest 1 cost st to each destination Shortest-path tree  Obtained by overlapping Src Dest 2 the individu idual al shortest 2 paths 2 1 Dest 1 20

Broa Br oadc dcast/Multic ast/Multicast ast Cla lassification ssification Broadcast Multicast One tree Shared tree Mesh per source (core-based tree) Single Multiple Minimize Minimize core core total cost cost to each node (Steiner tree) (e.g., Dijkstra) 21

Wi Wire reless less Mu Mult lticas icast t Adv Advan antag tage Wires  Locally distributing a packet to n neighbors  n times the cost of a unicast packet  Wireless: sending to n neighbors can incur costs  = tx to a single neighbor – if receive costs are ignored = One tx, n rx – if receives are correctly tuned = send n unicasts – if multicast not supported by MAC If local multicast is cheaper, then wireles ess  mul ulticast cast advantage ntage is present Can be assumed realistically  22

Ste teine iner r Tr Tree App Appro roximations imations Computing Steiner tree is NP complete  A simple approximation   Pick some arbitrary order of all destination nodes + source node  Successively add these nodes to the tree For every next node, construct a shortest path to  some other node already on the tree  Performs reasonably well in practice 23

Ste teine iner r Tr Tree App Appro roximations imations Takahashi Matsuyama heuristic   Similar, but let algorithm decide which is the next node to be added  Start with source node, add that destination node to the tree which has shortest path  Iterate, picking that destination node which has the shortest path to some node already on the tree Problem: Wireless multicast advantage not  exploited!  And does not really fit to the Steiner tree formulation 24

Br Broadcas oadcast t In Incr creme mental ntal Powe wer Or BIP  Exploits multicast wireless advantage   Goal: use as little transmission power as possible Based on Prim's MST algorithm  Once a node transmits and reaches some  neighbors, it becomes cheaper to reach additional neighbors 25

BI BIP – Ex Exam ample ple Round 1: Round 2: Round 3: A A A 5 4 2 3 3 3 S B B S (3) B S (1) 1 10 9 7 3 2 7 7 7 1 1 1 D D D Round 4: Round 5: C A C A C 2 3 3 S (3) B S (5) B 7 10 6 7 D D C (1) C (1) 26

Multic Mu lticast ast In Incr cremental mental Powe wer Or MIP  Start with broadcast tree construction, then  prune unnecessary edges out of the tree A A 3 3 S B S B 10 10 7 7 D D C C 27

Mesh-Based Me Based Mu Mult lticas icast Example – ODMRP (On-Demand Multicast  Routing Protocol) Sender NextHop H H Sender NextHop C A H C H D E B Sender NextHop G F H D I 28