MASTER THESIS SCALABLE ROUTING MECHANISM IN AD HOC ... 05-06... · Espoo, 09.05.2006 Tuna Vardar 5...
Transcript of MASTER THESIS SCALABLE ROUTING MECHANISM IN AD HOC ... 05-06... · Espoo, 09.05.2006 Tuna Vardar 5...
TKK Networking Laboratory
MASTER THESIS
SCALABLE ROUTING MECHANISMIN AD HOC NETWORKS
TUNA VARDAR
09.05.2006Espoo, Finland
Espoo, 09.05.2006 Tuna Vardar 2
TKK Networking Laboratory
Thesis Topic: Scalable Routing Mechanism in Ad Hoc networks
Thesis Supervisor: Prof. Raimo KantolaThesis Instructor: Lic. Sc. Jose Costa-RequenaThesis Workplace: MobileMan Project,
Networking Laboratory, ECE Department, TKK
Professorship: T-110, Telecommunications and Multimedia Laboratory, CSE Department, TKK
Espoo, 09.05.2006 Tuna Vardar 3
TKK Networking Laboratory
Outline
● Introduction● Objective● SARP simulation analysis● SARP implementation● SARP testing● Conclusions and future work
Espoo, 09.05.2006 Tuna Vardar 4
TKK Networking Laboratory
Introduction
● Ad Hoc networks● An emerging technology● There is no commercial demand
● But they are perfectly suited for disaster scenarios, search-and-rescue operations
● Nodes create Ad Hoc network dynamically without any fixed infrastructure
● Every node in the Ad Hoc network act as a router● As nodes move around, routes to other nodes in the
Ad Hoc network need to be discovered and maintained
● All nodes cooperate in carrying network traffic
Espoo, 09.05.2006 Tuna Vardar 5
TKK Networking Laboratory
Introduction
● Routing● Fundamental aspect of network establishment● The process of exchanging route information from one
node to other in a network● Ad Hoc networks routing
● It is a problem because there is no fixed infrastructure● Existing Internet routing protocols do not work very
well in Ad Hoc networks● Because they are designed considering that network has a fixed
infrastructure● Power constraints
● Low powered nodes do not produce ideal conditions for better route discovery and maintenance
Espoo, 09.05.2006 Tuna Vardar 6
TKK Networking Laboratory
Introduction
● Ad Hoc network routing protocols● Proactive routing protocols
● Maintain routing tables filled with the whereabouts of the other nodes● Use routing algorithms to periodically exchange link information● Consume more power and bandwidth● OLSR, DSDV
● On-demand routing protocols● Form the routes between nodes when needed● Consume less power, but still consumes unnecessary amounts of bandwidth● AODV, DSR, TORA
● Hybrid routing protocols● Combine advantages of proactive and on-demand routing protocols, thus
obtaining better performance● However, consume more memory and power, thus special precautions are
necessary on the nodes● ZRP
Espoo, 09.05.2006 Tuna Vardar 7
TKK Networking Laboratory
Objective
● To propose a new routing mechanism for Ad Hoc networks
● Scalable Ad Hoc Routing Protocol (SARP)● Hybrid solution
● Combines on-demand and proactive routing protocols in the same node
● Two routing protocols can share eachother's routing information
● Power constraints of the node are checked periodically to decide activation of proactive routing protocol or not for more efficient power consumption
● Supports external QoS module (optional)
Espoo, 09.05.2006 Tuna Vardar 8
TKK Networking Laboratory
Objective
● SARP● Main target is to reduce the time spent in the route
discovery● Proactive routing stores link information
● It eliminates time spent in the route discovery by on-demand routing protocols
● This operation consumes more power but can be managed● If the node is able to decide activation or deactivation of proactive routing,
excess power consumption can be managed and routing efficiency is increased● Nodes are classified into two groups
● Smart nodes● Nodes running simultaneously on-demand and proactive routing protocols
● Ordinary nodes● Nodes running only on-demand routing protocols
Espoo, 09.05.2006 Tuna Vardar 9
TKK Networking Laboratory
Objective
● SARP● Each smart nodes decides to be smart or ordinary node
by checking its power level● If node's battery power level is greater than predefined power
threshold, node is “Smart”.● Smart nodes will communicate between them using
proactive routing protocol● This will create so called a virtual backbone of smart nodes
● Smart nodes will communicate with ordinary nodes using on-demand routing protocol● On-demand routing protocol should be always active on the
smart nodes● Ordinary nodes will communicate between them and
with smart nodes using on-demand routing protocol
Espoo, 09.05.2006 Tuna Vardar 10
TKK Networking Laboratory
SARP Simulation Analysis
● Performance comparison of SARP with state of the art routing protocols● Necessary for correctly verification of SARP● Simulation environment
● NS-2 running in Linux● SARP implemented for NS-2 using C++
● Simulation results● Data loss● Delay● Routing data size● Throughput● Overhead
Espoo, 09.05.2006 Tuna Vardar 11
0 30 60 120 300 600 9000
100
200
300
400
500
600
700
800
900
1000
Delay
DSDVDSRTORAOLSRAODVSARP-10SARP-15SARP-25
Pauses between node movements (seconds)
Del
ay (m
ilisec
onds
)
0 30 60 120 300 600 900
0
3
5
8
10
13
15
18
20
23
25
28
30
33
35
38
Data Loss
DSDVDSRTORAOLSRAODVSARP-10SARP-15SARP-25
Pauses between node movements (seconds)
Dat
a lo
ss (p
erce
ntag
e) TKK Networking Laboratory
SARP Simulation Analysis - I
Simulation parameters
● SARP, AODV, OLSR, DSR, DSDV, TORA
● 50 nodes● 5, 10, 25 smart nodes for SARP● 1500 meters width, 300 meter height● 7 different mobility levels at X-axis● Results are at Y-axis● 30 CBR traffic connection
– 64 bytes UDP packets– Duration: 12,5 seconds– Rate: 8 packets/second
● Radio range: 250 meters
Espoo, 09.05.2006 Tuna Vardar 12
0 30 60 120 300 600 9009,5
10,010,5
11,011,5
12,012,5
13,013,514,0
14,515,0
15,5
Throughput
DSDVDSRTORAOLSRAODVSARP-10SARP-15SARP-25
Pauses between node movements (seconds)
Thro
ughp
ut (k
iloby
tes/
seco
nd)
0 30 60 120 300 600 9000
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
Routing Data Size
DSDVDSRTORAOLSRAODVSARP-10SARP-15SARP-25
Pauses between node movements (seconds)
Rou
ting
data
siz
e (k
iloby
tes)
TKK Networking Laboratory
SARP Simulation Analysis - I
Simulation parameters
● SARP, AODV, OLSR, DSR, DSDV, TORA
● 50 nodes● 5, 10, 25 smart nodes for SARP● 1500 meters width, 300 meter height● 7 different mobility levels at X-axis● Results are at Y-axis● 30 CBR traffic connection
– 64 bytes UDP packets– Duration: 12,5 seconds– Rate: 8 packets/second
● Radio range: 250 meters
Espoo, 09.05.2006 Tuna Vardar 13
TKK Networking Laboratory
SARP Simulation Analysis - II
Effect of node densitySimulation parameters● SARP, AODV, OLSR● 25,50,100 nodes● 10%, 30%, 60% smart nodes for SARP● 1500 meters width, 300 meter height
● 7 different mobility levels at X-axis● Results are at Y-axis● 30 TCP traffic connections
– 64 bytes TCP packets– Duration: 20 seconds– FTP is used at application layer
● Radio range: 250 meters
Espoo, 09.05.2006 Tuna Vardar 14
TKK Networking Laboratory
SARP Simulation Analysis - II
Effect of node densitySimulation parameters● SARP, AODV, OLSR● 25,50,100 nodes● 10%, 30%, 60% smart nodes for SARP● 1500 meters width, 300 meter height
● 7 different mobility levels at X-axis● Results are at Y-axis● 30 TCP traffic connections
– 64 bytes TCP packets– Duration: 20 seconds– FTP is used at application layer
● Radio range: 250 meters
Espoo, 09.05.2006 Tuna Vardar 15
TKK Networking Laboratory
SARP Simulation Analysis - II
Effect of node densitySimulation parameters● SARP, AODV, OLSR● 25,50,100 nodes● 10%, 30%, 60% smart nodes for SARP● 1500 meters width, 300 meter height
● 7 different mobility levels at X-axis● Results are at Y-axis● 30 TCP traffic connections
– 64 bytes TCP packets– Duration: 20 seconds– FTP is used at application layer
● Radio range: 250 meters
Espoo, 09.05.2006 Tuna Vardar 16
0 30 60 120 300 600 9000
5
10
15
20
25
30
35
40
45
Data Loss Comparison
AODV (50 s) OLSR (50 s)SARP(15 smart, 50s)OLSR (12.5 s)AODV (12.5 s)SARP(15 smart,12.5 s)
Pauses between node movements (seconds)
Dat
a lo
ss (p
erce
ntag
e)
0 30 60 120 300 600 9000
100
200
300
400
500
600
700
800
900
1000
Delay
AODV (50 s) OLSR (50 s)SARP(15 smart, 50s)OLSR (12.5 s)AODV (12.5 s)SARP(15 smart,12.5 s)
Pauses between node movements (seconds)
Del
ay (m
ilise
cond
s) TKK Networking Laboratory
SARP Simulation Analysis - III
Simulation parameters
● SARP, AODV, OLSR● 50 nodes● 5, 10, 25 smart nodes for SARP● 1500 meters width, 300 meter height● 7 different mobility levels at X-axis● Results are at Y-axis● 30 CBR traffic connection
– 65 bytes UDP packets– Duration: 50 seconds– Rate: 8 packets/second
● Radio range: 250 meters● Effect of traffic congestion
– Continuous lines show results for third simulation analysis
– Dashed lines show results for first simulation analysis
Espoo, 09.05.2006 Tuna Vardar 17
0 30 60 120 300 600 900
0
250
500
750
1000
1250
1500
1750
2000
2250
2500
Routing Data Size
AODV (50 s) OLSR (50 s)SARP(15 smart, 50s)OLSR (12.5 s)AODV (12.5 s)SARP(15 smart,12.5 s)
Pauses between node movements (seconds)
Rou
ting
data
siz
e (k
iloby
tes)
0 30 60 120 300 600 9008,59,09,5
10,010,511,011,512,012,513,013,514,014,515,015,516,0
Throughput
AODV (50 s) OLSR (50 s)SARP(15 smart, 50s)OLSR (12.5 s)AODV (12.5 s)SARP(15 smart,12.5 s)
Pauses between node movements (seconds)
Thro
ughp
ut (k
iloby
tes/
seco
nd)
TKK Networking Laboratory
SARP Simulation Analysis - III
Simulation parameters
● SARP, AODV, OLSR● 50 nodes● 5, 10, 25 smart nodes for SARP● 1500 meters width, 300 meter height● 7 different mobility levels at X-axis● Results are at Y-axis● 30 CBR traffic connection
– 65 bytes UDP packets– Duration: 50 seconds– Rate: 8 packets/second
● Radio range: 250 meters● Effect of traffic congestion
– Continuous lines show results for third simulation analysis
– Dashed lines show results for first simulation analysis
Espoo, 09.05.2006 Tuna Vardar 18
TKK Networking Laboratory
SARP Simulation Analysis
● Conclusions● SARP provides lower delay and data loss than OLSR,
but similar to AODV● SARP uses less data for routing than OLSR, but slightly
more than AODV● When amount of smart nodes increases, it does not
have positive effect in the performance● Best performance is observed when 10-30 percent of nodes is
smart● SARP performance is similar to AODV
● SARP decrease data loss slightly in some cases● SARP decreases delay● SARP increases throughput slightly in some cases● SARP increases routing data size
Espoo, 09.05.2006 Tuna Vardar 19
TKK Networking Laboratory
SARP Implementation
● Combination of AODV and OLSR in the same memory address space
● AODV and OLSR existing functionality is kept unchanged
● 3 new modules added for SARP functionality● Node identification module added to AODV impl.
● Decides if the node is smart or ordinary● QoS integration module added to AODV impl.
● Initiated by node identification module● Supports integration of SARP with external QoS module
(optional)● Routing table merging module added to OLSR impl.
● Merges routes from AODV routing cache into OLSR routing table● Merges routes from OLSR routing table into AODV routing cache
Espoo, 09.05.2006 Tuna Vardar 20
TKK Networking Laboratory
SARP Implementation
● Implementation is done using C language● AODV-UU 0.9● Unik OLSR 0.4.4
● Built for iPAQ PDA devices running Familiar Linux
● ARM cross compiler● Familiar Linux kernel source code
Espoo, 09.05.2006 Tuna Vardar 21
TKK Networking Laboratory
SARP Implementation
● Node identification module
Espoo, 09.05.2006 Tuna Vardar 22
TKK Networking Laboratory
SARP Implementation
● QoS integration module
Espoo, 09.05.2006 Tuna Vardar 23
TKK Networking Laboratory
SARP Implementation
● Routing table merging module
Espoo, 09.05.2006 Tuna Vardar 24
TKK Networking Laboratory
SARP testing
● Four different test cases● Tested on iPAQ PDAs running Familiar Linux● Test case I
● Communication between smart node and ordinary node● Communication was totally handled by AODV● Initial route discovery was 2-3 seconds
● Test number 1: 2173 ms● Test number 2: 1968 ms● Test number 3: 1725 ms● Test number 4: 2483 ms● Test number 5: 2268 ms
Espoo, 09.05.2006 Tuna Vardar 25
TKK Networking Laboratory
SARP testing
● Test case II● Communication between smart node and smart node● Communication was handled by AODV● OLSR module was running and updating AODV routing
table● Initial route discovery was 2-3 seconds
● Test number 1: 2423 ms● Test number 2: 2489 ms● Test number 3: 2654 ms● Test number 4: 2543 ms● Test number 5: 2802 ms
Espoo, 09.05.2006 Tuna Vardar 26
TKK Networking Laboratory
SARP testing
● Test case III● Communication between two smart nodes and an
ordinary node● Communication was handled by AODV and OLSR● OLSR module was running and updating AODV routing
table● Initial route discovery from Node 1 to Node 3 takes 3-4
seconds● Test number 1: 3372 ms● Test number 2: 3464 ms● Test number 3: 3781 ms● Test number 4: 3267 ms● Test number 5: 3466 ms
Espoo, 09.05.2006 Tuna Vardar 27
TKK Networking Laboratory
SARP testing
● Test case IV● Communication between two smart nodes and two
ordinary nodes● Communication was handled by AODV● OLSR module was running and updating AODV routing
table● Initial route discovery from node 4 to node 3 takes 4-5
seconds● Test number 1: 4835 ms● Test number 2: 4904 ms● Test number 3: 5067 ms● Test number 4: 5384 ms● Test number 5: 5130 ms
Espoo, 09.05.2006 Tuna Vardar 28
TKK Networking Laboratory
Conclusions and future work
● Good results for SARP in simulations● Increase in number of smart nodes do not increase
the overall performance● Up to the simulation results, SARP performance is
similar to AODV, even better in same cases● Real-time tests show SARP implementation as stable enough
● Real-time tests show AODV and OLSR inside SARP exchanging routes● This decreases time spent for initial route discovery● Benefit of SARP may be seen better in larger size Ad
Hoc networks as it is shown by simulations● Real-time tests in a bigger network is good to do in the future