You are on page 1of 24

Ad Hoc Networks:#

Architectures & Protocols

Dr Ljiljana Simić
iNETS, RWTH Aachen University
SS2016

Important  Note:  
These   course   notes   may   contain   some   copyrighted  
material.   The   copyright   of   this   material   covers   its  
use   in   class   and   for   educa7onal   purposes,   but   you  
are   not   allowed   to   distribute   this   course   material  
freely.   Under   the   code   of   appropriate   use,   please  
refrain   from   uploading   the   provided   source   files   or  
documents  to  any  publicly  accessible  system  outside  
RWTH  Aachen  University  without  prior  permission.  

1
Announcement: Extra Exam Day
 
 
§  due  to  unexpectedly  high  exam  registra7on  numbers,  we  
must  add  an  extra  day  to  the  Ad  Hoc  oral  exam  period:  
Friday,  August  5th    

§  if  you  have  a  CLASH  with  another  exam  on  5th  August,  
you  MUST  send  Fei  an  email  reporDng  the  clash  by  
Thursday  23rd  June.  

§  we  will  then  announce    the  oral  exam  schedule  


(i.e.  assigned  7me  slot  per  student)  

– Routing for Ad Hoc Networks (IV)


#

2
Cluster-based routing
§  examples  of  cluster-­‐based  schemes:  
§  [Gerla  and  Tsai,  “MulDcluster,  mobile,  mulDmedia  radio  network,”  
Wireless  Networks,  1995]  
§  [Krishna,  Vaidya,  ChaQerjee  and  Pradhan,  “A  cluster-­‐based  approach  for  
rouDng  in  dynamic  networks,”  ACM  SIGCOMM  CCR,  1997.]  
§  [Amis,  Prakash,  Vuong  and    Huynh,  “Max-­‐min  d-­‐cluster  formaDon  in  
wireless  ad  hoc  networks,”    in  Proc.  INFOCOM,  2000.]  
 

§  typically  a  leader  is  elected  for  each  cluster  of  nodes  
§  leader  oXen  has  some  special  responsibili7es  
§  schemes  may  differ  in:  
§  how  clusters  are  determined  
§  the  way  cluster  head  (leader)  is  chosen  
§  duDes  assigned  to  the  cluster  head  
 
 

Proactive routing protocols


§  most  of  the  schemes  discussed  so  far  are  reac7ve  

§  proacDve  schemes  based  on  distance-­‐vector  and    


link-­‐state  mechanisms  have  also  been  proposed  for  MANET  

Link-State vs. Distance-Vector

3
Proactive routing protocols
§  most  of  the  schemes  discussed  so  far  are  reac7ve  

§  proacDve  schemes  based  on  distance-­‐vector  and    


link-­‐state  mechanisms  have  also  been  proposed  for  MANET  

§  distance-­‐vector  rouDng:  each  node  …  


§  broadcasts  to  neighbours  its  view  of  distance  to  all  other  nodes  
§  then  computes  shortest  path  to  each  other  node  based  on  
informa7on  adver7sed  by  its  neighbours  

§  link-­‐state  rouDng:  each  node  …  


§  broadcasts  to  all  other  nodes  its  view  of  status  of  its  neighbour  links  
§  then  computes  shortest  path  to  each  other  node  based  on  complete  
picture  of  network  from  most  recent  link  informa7on  from  all  nodes  

Distance-vector routing: DSDV


(review)
[Perkins  and  Bhagwat,  “Highly  dynamic  des7na7on-­‐sequenced  distance-­‐vector  
rou7ng  (DSDV)  for  mobile  computers,”  in  Proc.  ACM  SIGCOMM,  1994.]  
 

§  each  node  maintains  a  rouDng  table  which  stores  


§  next  hop  towards  each  des7na7on  
§  cost  metric  for  the  path  to  each  des7na7on  
§  des7na7on  sequence  number  created  by  des7na7on  itself  
§  sequence  numbers  used  to  avoid  forma7on  of  loops  

§  each  node  periodically  forwards  the  rouDng  table  to  its  
neighbours  
§  each  node  increments  and  appends  its  sequence  number  when  
sending  its  local  rou7ng  table  
§  sequence  number  aQached  to  route  entries  created  for  this  node  

4
DSDV (review)

§  assume  node  X  receives  rou7ng  informa7on  from  Y  


about  a  route  to  node  Z  

X   Y   Z  

§  let  S(X)  =  des7na7on  sequence  number  for  node  Z  as  


stored  at  node  X  
 

§  let  S(Y)  =  des7na7on  sequence  number  for  node  Z  as  


sent  by  node  Y  with  its  rou7ng  table  to  node  X  

DSDV (review)
§  node  X  takes  the  following  steps:  

X   Y   Z  
 
§  if    S(X)  >  S(Y),  
   X  ignores  the  rou7ng  informa7on  received  from  Y    
 

§  if  S(X)  =  S(Y)  &  cost  of  going  through  Y  is  smaller  than  the  route  
known  to  X,  
   then  X  sets  Y  as  the  next  hop  to  Z  

§  if  S(X)  <  S(Y),  


   X  sets  Y  as  next  hop  to  Z  and  S(X)  updated  to  equal  S(Y)  

5
Link-state routing
§  each  node:  
§  periodically  floods  status  of  its  links  
§  re-­‐broadcasts  link  state  informa7on  received  from  its  
neighbours  
§  keeps  track  of  link  state  informa7on  received  from  other  
nodes  
§  uses  above  informa7on  to  determine  next  hop  to  each  
des7na7on  

Optimized Link State Routing (OLSR)


 
 
[Jacquet,  et  al,  “Op7mized  link  state  rou7ng  protocol  for  ad  hoc  networks,”  in  
 
Proc.  INMIC,  2001.]  

§  overhead  of  flooding  link  state  informa7on  is  reduced  by  
requiring  fewer  nodes  to  forward  the  informa7on  

§  relies  on  selecDon  of  mulDpoint  relays  &  calculates  its  
routes  to  all  known  des7na7ons  through  these  nodes  
 

6
Optimized Link State Routing (OLSR)
[Jacquet,  et  al,  “Op7mized  link  state  rou7ng  protocol  for  ad  hoc  networks,”  in  
 
Proc.  INMIC,  2001.]  

§  O/H  of  flooding  link  state  informa7on  reduced  by  selecDon  
of  mulDpoint  relays  

OLSR
§  a  broadcast  from  node  X  is  only  forwarded  by  its  
mulDpoint  relays  
 

§  node  X  selects  its  mul7point  relay  set  among  its  one-­‐hop  
neighbours,  such  that  the  set  covers  all  two-­‐hop  neighbours    
§  i.e.  such  that  each  two-­‐hop  neighbour  of  X  is  a  one-­‐hop  
neighbour  of  at  least  one  mul7point  relay  of  X  
§  MPR  set  need  not  be  op7mal  
§  selec7on  of  op7mal  MRP  set  is  NP-­‐complete,  heuris7c  in  
[Jacquet2001]  
§  but  the  smaller  MRP  sets  are,  the  more  efficient  OLSR  is  
 

§  each  node  transmits  its  neighbour  list  in  periodic  beacons,  
so  that  all  nodes  can  know  their  2-­‐hop  neighbours,  in  order  to  
choose  the  mulDpoint  relays  

7
OLSR
 
§  nodes  C  and  E  are  mul7point  relays  of  node  A  

B   F   J  

A   E   H  
C   K  
G  
D  

node  that  has  link-­‐state  informa7on  broadcasted  from  A  

OLSR
 
§  nodes  C  and  E  forward  informa7on  received  from  A  

B   F   J  

A   E   H  
C   K  
G  
D  

node  that  has  link-­‐state  informa7on  broadcasted  from  A  

8
OLSR
§  nodes  E  and  K  are  mul7point  relays  for  node  H  
§  node  K  forwards  informa7on  received  from  H  
§  node  E  has  already  forwarded  the  same  informa7on  once  

B   F   J  

A   E   H  
C   K  
G  
D  

node  that  has  link-­‐state  informa7on  broadcasted  from  A  

OLSR
§  floods  informa7on  through  the  mul7point  relays  
§  the  flood  itself  is  for  links  connecDng  nodes  to  respecDve  
mulDpoint  relays  
§  routes  used  by  OLSR  only  include  mulDpoint  relays  as  
intermediate  nodes    
 

§  each  node  in  network  periodically  broadcasts  informaDon  about  


its  one-­‐hop  neighbours  which  have  selected  it  as  a  mulD-­‐point  
relay  
§  upon  receipt  of  this  info,  each  node  calculates  and  updates  its  
routes  to  each  known  des7na7on  
§  route  is  a  sequence  of  hops  through  the  mul7point  relays  from  
source  to  des7na7on  

9
OLSR

OLSR
§  advantage  vs.  distance-­‐vector  protocols:  
reduces  rou7ng  O/H  &  number  of  broadcasts  
 à  low  connec7on  setup  7me  &reduced  control  O/H  

10
Hybrid routing protocols
remember  …  

proactive vs. reactive


§  maintains  routes  to  every  other   §  maintains  routes  to  only  those  nodes  
node  in  the  network   which  are  needed  
§  regular  rouDng  updates  impose   §  cost  of  finding  routes  is  expensive  
large  overhead   since  flooding  is  involved  
§  no  latency  in  route  discovery,  i.e.   §  might  be  delay  before  transmiong  
data  can  be  sent  immediately   data  
§  most  rou7ng  informaDon  might   §  cache  informa7on  from  other  nodes’  
never  be  used   transmissions    
§  suitable  for  high  traffic  networks   §  good  for  low/medium  traffic  
networks  

Zone Routing Protocol (ZRP)


 
[Haas,  “A  new  rou7ng  protocol  for  the  reconfigurable  wireless  networks,”  in  
 
Proc.  IEEE  ICUPC,  1997.]  
 
§  ZRP  is  a  hybrid  rou7ng  protocol  which  combines:  
§  proac7ve  protocol  –  proac7vely  updates  network  state  
and  maintains  route  regardless  of  whether  any  data  
traffic  exists  or  not  

§  reac7ve  protocol  –    only  determines  route  to  a  


des7na7on  if  there  is  some  data  to  be  sent  to  the  
des7na7on  

11
ZRP

§  all  nodes  within  hop  distance  of  at  most  d  from  a  node  X  
are  said  to  be  in  the  rouDng  zone  of  node  X  

§  all  nodes  at  hop  distance  exactly  d  are  said  to  be  
peripheral  nodes  of  node  X’s  rou7ng  zone  

ZRP
§  intra-­‐zone  rouDng:  proac7vely  maintain  state  informa7on  
for  links  within  a  short  distance  from  any  given  node  
§  routes  to  nodes  within  short  distance  are  thus  
maintained  proacDvely  
(using,  say,  link  state  or  distance  vector  protocol)  

§  inter-­‐zone  rouDng:  use  a  route  discovery  protocol  for  


determining  routes  to  far  away  nodes  
§  route  discovery  is  similar  to  DSR  with  the  excep7on  
that  route  requests  are  propagated  via  peripheral  
nodes  

12
ZRP: example (zone radius, d = 2)

B  

S  
A   C  
D  
E  
F  

peripheral  nodes  in  


rou7ng  zone  of  node  S    

ZRP: example (zone radius, d = 2)

§  S  performs  route  


discovery  for  D  
B  

S  
A   C  
D  
E  
F  

denotes  route  request  

13
ZRP: example (d = 2)

§  S  performs  route  


discovery  for  D  
B  
S  
A   C  
D  
E  
F  

§  E  knows  route  from  E  to  D,    so  


route  request  need  not  be  
denotes  route  reply   forwarded  to  D  from  E  

ZRP: example (d = 2)

§  S  performs  route  


discovery  for  D  
B  
S  
A   C  
D  
E  
F  

denotes  route  taken  by  Data  

14
ZRP
§  Advantages:  
§  combines  best  features  of  proac7ve  and  reac7ve  rou7ng  
§  reduces  control  O/H  compared  to  RREQ  flooding  in  
on-­‐demand  approaches  &  periodic  flooding  of  rou7ng  
informa7on  packets  on  table-­‐driven  approaches  

§  Disadvantages:  
§  in  absence  of  query  control  (ensuring  redundant  or  duplicate  
RREQ  not  forwarded),  produces  higher  control  O/H  
§  e.g.  due  to  large  overlapping  of  nodes’  rou7ng  zones  
§  decision  on  zone  radius  significantly  impacts  performance  
§  can  be  difficult  to  assure  route  stability  

Landmark Ad hoc Routing (LANMAR)#


with group mobility
[Pei,  Gerla  and  Hong,  “LANMAR:  landmark  rou7ng  for  large  scale  wireless  ad  hoc  
networks  with  group  mobility,”  in  Proc.  ACM  MobiHoc,  2000.]
 
   
§  landmark  node  is  elected  for  a  group  of  nodes  that  are  
likely  to  move  together  

§  scope  is  defined  such  that  each  node  would  typically  be  
within  the  scope  of  its  landmark  node  

§  each  node  propagates:  link-­‐state  informa7on  corresponding  


only  to  nodes  within  its  scope  &  distance-­‐vector  informa7on  for  
all  landmark  nodes  
§  combina7on  of  link-­‐state  and  distance-­‐vector  
§  distance-­‐vector  used  for  landmark  nodes  outside  the  scope  
§  no  state  informa7on  for  non-­‐landmark  nodes  outside  scope  maintained  

15
LANMAR: routing to nodes within scope
§  assume  node  C  is  within  scope  of  node  A  
H  

G  
C   D  

A   B   E  
F  

§  rouDng  from  A  to  C:  node  A  can  determine  next  hop  to  
node  C  using  the  available  link  state  informaDon  

LANMAR: routing to nodes outside scope


§  let  H  be  the  landmark  node  for  node  F  
H  

G  
C   D  

A   B   E  
F  

§  rouDng  from  node  A  to  F,  which  is  outside  A’s  scope:  
§  node  A  somehow  knows  that  H  is  the  landmark  for  F  
§  node  A  can  determine  next  hop  to  node  H  using  the  
available  distance  vector  informaDon  

16
LANMAR: routing to nodes outside scope
§  node  D  is  within  scope  of  node  F  
H  

G  
C   D  

A   B   E  
F  

§  node  D  can  determine  next  hop  to  node  F  using  


link  state  informaDon  
§  packet  for  F  may  never  reach  the  landmark  node  H,  
even  though  iniDally  node  A  sends  it  towards  H  

LANMAR
§  uses  node  iden7fiers  as  landmarks  

§  Anchored  Geodesic  Scheme  [LeBoudec,  2000]  


uses  geographical  regions  as  landmarks  

17
Geodesic routing: without anchors
[Blazevic  ,  Giordano  ,  Le  Boudec  ,  “Self-­‐organized  wide-­‐area  rou7ng,”  
in  Proc.  SCI/ISAS,  2000.]  
[Hubaux,  Le  Boudec,  Giordano  ,  Hamdi,  Blazevic,  BuQyan  and  Vojinovic,  

“Towards  mobile  ad-­‐hoc  WANs:  terminodes,”  in  Proc.  WCNC,  2000.]    
§  intra-­‐zone  rouDng:  
each  node  somehow  keeps  track  of  routes  to  nodes  within  its  zone    
&  records  physical  loca7ons  of  nodes  on  its  zone  boundary  
§  inter-­‐zone  rouDng:  
when  packet  is  to  be  routed  to  someone  outside  zone,  it  is  sent  to  
a  zone-­‐boundary  node  in  the  direcDon  of  the  desDnaDon  
§  packet  is  forwarded  in  this  manner  un7l  it  reaches  someone  within  
the  des7na7on’s  zone  
§  this  node  then  uses  intra-­‐zone  rou7ng  to  deliver  the  packet  
§  similar  to  the  GEDIR  protocol  [Lin,  1998]  

Anchored geodesic routing


§  anchors  can  be  used  to  go  around  connec7vity  holes  

§  anchors  are  physical  loca7ons/areas  

§  route  may  be  specified  as  a  series  of  intermediate  


physical  areas  to  be  traversed  to  reach  the  des7na7on  

B  

A  

18
Ensuring routes exist
§  protocols  discussed  thus  far  find/maintain  a  route  
provided  it  exists  

§  some  protocols  aeempt  to  ensure  that  a  route  exists  by:  
§  power  control  
§  limi7ng  movement  of  hosts  or  forcing  them  to  take  
detours  

Ensuring routes exist: power control


§  protocols  discussed  thus  far  find  a  route,  on  a  given  network  
topology  

§  some  researchers  propose  controlling  network  topology  by  


transmission  power  control  to  yield  network  proper7es  
which  may  be  desirable,  e.g.[Ramanathan,  INFOCOM  2000]    
§  such  approaches  can  significantly  impact  performance  at  
several  layers  of  protocol  stack  

§  [WaQenhofer,  INFOCOM  2001]    provides  a  distributed  


mechanism  for  power  control  which  allows  for  local  
decisions,  but  guarantees  global  connec7vity  
§  each  node  uses  a  power  level  that  ensures  that  the  node  has  at  least  
one  neighbor  in  each  cone  with  angle  2π/3  

19
Even more routing protocols …
 

§  plenty  of  other  rou7ng  protocols:  


believe  it  or  not,  discussion  here  is  far  from  exhaus7ve!  

§  many  of  the  exis7ng  protocols  could  poten7ally  be  


adapted  for  MANET  (some  have  already  been  adapted,  
as  discussed  earlier)  

§  let’s  look  at  a  few  more  varia7ons  …  

Power-aware routing
[Singh,  Woo  and  Raghavendra,  “Power-­‐aware  rou7ng  in  mobile  ad  hoc  networks,”  in  Proc.  
ACM  MobiCom,  2000.]  
[Chang,  “Energy  conserving  rou7ng  in  wireless  ad-­‐hoc  networks,”  in  Proc.  IEEE  INFOCOM,  
2000.]  
 

§  define  opDmizaDon  criteria  as  a  funcDon  of  energy  


consumpDon,  e.g.:  
§  minimize  energy  consumed  per  packet  
§  minimize  7me  to  network  par77on  due  to  energy  
deple7on  
§  maximize  node  life7me  (dura7on  before  a  node  fails  due  
to  energy  deple7on)  

20
Power-aware routing
§  assign  a  weight  to  each  link  

§  weight  of  a  link  may  be  a  func7on  of  energy  consumed  
when  transmifng  a  packet  on  that  link,  as  well  as  the  
residual  energy  level  
§  low  residual  energy  level  may  correspond  to  a  high  
cost  

§  prefer  a  route  with  the  smallest  aggregate  weight  

Power-aware routing
§  possible  modificaDon  to  DSR  to  make  it  power  aware    
(for  simplicity,  assume  no  route  caching):  
 

§  RREQ’s  aggregate  the  weights  of  all  traversed  links  


§  desDnaDon  responds  with  a  RREP  to  a  RREQ  if:  
§  it  is  the  first  RREQ  with  a  given  (“current”)  sequence  
number,  
OR  
§  its  weight  is  smaller  than  all  other  RREQs  received  
with  the  current  sequence  number  

21
Signal Stability Based Adaptive Routing
(SSA)
[Dube,  Rais,  Wang  and  Tripathi,  “Signal  stability-­‐based  adap7ve  rou7ng  (SSA)  for  ad  hoc  
mobile  networks  ,”IEEE  Personal  Commun.,  1997.]  

§  similar  to  DSR  


§  node  X  re-­‐broadcasts  a  RREQ  received  from  Y  only  if  the  
(X,Y)  link  is  deemed  to  have  a  strong  signal  stability  
§  signal  stability  is  evaluated  as  a  moving  average  of  signal  
strength  of  packets  received  on  the  link  in  recent  past  
§  alterna7ve  approach  would  be  to  assign  a  cost  as  a  func7on  
of  signal  stability  

Associativity Based Routing (ABR)#


networks,”  Wireless  Pers.  Comm.,  1997.]  
[Toh,“Associa7vity-­‐based  rou7ng  for  ad  hoc  mobile  

§  use  only  links  that  have  been  stable  for  some  minimum  
 
duraDon  
§  moDvaDon:  
§  if  a  link  has  been  stable  beyond  some  minimum  
threshold,  it  is  likely  to  be  stable  for  a  longer  interval  
§  if  it  has  not  been  stable  longer  than  the  threshold,  then  it  
may  soon  break  (could  be  a  transient  link)  

§  associaDon  stability  determined  for  each  link  


§  measures  dura7on  for  which  the  link  has  been  stable  

§  prefer  paths  with  high  aggregate  associa7on  stability  

22
Geographical Adaptive Fidelity (GAF)
[Xu,  Heidemann  and  Estrin,“Geography-­‐informed  energy  conserva7on  for  ad  hoc  rou7ng,”  
in  Proc.  ACM  MobiCom,  2001.]  

§  each  node  associates  itself  with  a  square  in  a  virtual  grid  

§  nodes  in  each  grid  square  coordinate  to  determine  who  
will  sleep  and  how  long  
 

Preemptive routing
[Goff  et  al.,  “Preemp7ve  rou7ng  in  ad  hoc  networks,”  in  Proc.  ACM  MobiCom,  2001.]  

§  add  some  proac7vity  to  reac7ve  rou7ng  protocols  such  as  
DSR  and  AODV  

§  route  discovery  iniDated  when  it  appears  that  an  acDve  
route  will  break  in  the  near  future  
§  e.g.  path  considered  likely  to  break  when  received  
packet  power  becomes  close  to  the  minimum  
detectable  power  

§  ini7a7ng  route  discover  before  exis7ng  route  breaks  


reduces  discovery  latency  

23
24

You might also like