simulating a mobile peer-to-peer network simo sibakov department of communications and networking...

19
SIMULATING A MOBILE PEER-TO-PEER SIMULATING A MOBILE PEER-TO-PEER NETWORK NETWORK Simo Sibakov Simo Sibakov Department of Communications and Networking (Comnet) Department of Communications and Networking (Comnet) Helsinki University of Technology Helsinki University of Technology Supervisor: Prof. Raimo Kantola Supervisor: Prof. Raimo Kantola

Upload: shannon-gregory

Post on 23-Dec-2015

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

SIMULATING A MOBILE PEER-TO-PEER SIMULATING A MOBILE PEER-TO-PEER NETWORKNETWORK

Simo SibakovSimo SibakovDepartment of Communications and Networking (Comnet)Department of Communications and Networking (Comnet)

Helsinki University of TechnologyHelsinki University of Technology

Supervisor: Prof. Raimo KantolaSupervisor: Prof. Raimo Kantola

Page 2: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

ContentsContents► BackgroundBackground► Protocol introductionsProtocol introductions

SIP, P2PSIP, RELOADSIP, P2PSIP, RELOAD

► Overlay ConceptOverlay Concept► DHTDHT

conceptconcept DHT algorithms Chord and KademliaDHT algorithms Chord and Kademlia

► Routing modesRouting modes► OverSim simulatorOverSim simulator► Input parametersInput parameters► Output parametersOutput parameters► ResultsResults► ConclusionsConclusions

Page 3: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

Background of the studyBackground of the study► The task was to:The task was to:

simulate a P2PSIP network using RELOAD as its peer simulate a P2PSIP network using RELOAD as its peer protocolprotocol

evaluate P2PSIP’s applicability for mobile telephone evaluate P2PSIP’s applicability for mobile telephone networksnetworks

► The study included:The study included: Writing a C++ class for the simulation program to model Writing a C++ class for the simulation program to model

the RELOAD messagesthe RELOAD messages Making other necessary additions to the simulation Making other necessary additions to the simulation

program code.program code. Collecting and analyzing the resultsCollecting and analyzing the results

Page 4: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

SIP (Session Initiation Protocol)SIP (Session Initiation Protocol)

► SIP is a signaling protocol used for setting up SIP is a signaling protocol used for setting up multimedia sessionsmultimedia sessions

► SIP supports user mobility and localizationSIP supports user mobility and localization► SIP is a traditional client/server protocol dependent SIP is a traditional client/server protocol dependent

on server elementson server elements

Page 5: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

LocationService

1 REGISTER

5 Response

2 Store

3 INVITE

4 QueryProxyRegistrar

6 INVITE

UA

UA

domain.com

sip.domain.com

[email protected]

[email protected]

[email protected]

Alice@work

John@cafeteria

(public URI)

(SIP URI)

SIP MESSAGE

OTHER PROTOCOL

Page 6: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

P2PSIP (Peer-to-Peer SIP)P2PSIP (Peer-to-Peer SIP)► An alternative solution to the session establishmentAn alternative solution to the session establishment► Provides the same localization service as SIP but without Provides the same localization service as SIP but without

server elementsserver elements► Every node (peer) has equal functions and responsibilities Every node (peer) has equal functions and responsibilities

for data storage and message routingfor data storage and message routing

RELOAD (Resource LOcation And Delivery)RELOAD (Resource LOcation And Delivery)► The peer protocol of P2PSIPThe peer protocol of P2PSIP► RELOAD is used for inter-peer communications in the RELOAD is used for inter-peer communications in the

P2PSIP overlayP2PSIP overlay

Page 7: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

DHT (Distributed Hash Table)DHT (Distributed Hash Table)

► Distributes the data and query load evenly to Distributes the data and query load evenly to all nodes in the overlayall nodes in the overlay

► Divides the keyspace between the Divides the keyspace between the participating nodesparticipating nodes

► Each node has an identifier (node ID)Each node has an identifier (node ID)► Stored data elements also have identifiers Stored data elements also have identifiers

(resource ID)(resource ID)► Two DHT algorithms, Chord and Kademlia, are Two DHT algorithms, Chord and Kademlia, are

used in this studyused in this study► DHT algorithms define the logical location of DHT algorithms define the logical location of

the nodes in the overlaythe nodes in the overlay

Page 8: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

Physical links

Logical links

UNDERLAYING NETWORK

OVERLAY NETWORK

Overlay concept

Page 9: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

ChordChord► Nodes are logically arranged in the identifier Nodes are logically arranged in the identifier

circlecircle► Keys are assigned to the node whose ID succeeds Keys are assigned to the node whose ID succeeds

the ID of the keythe ID of the key► Chord has two independent routing tables: Chord has two independent routing tables:

successor list and finger tablesuccessor list and finger table

Page 10: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

KademliaKademlia► Nodes are treated as leaves of a binary tree where Nodes are treated as leaves of a binary tree where

nodes are located according to the shortest unique nodes are located according to the shortest unique prefix of the node ID.prefix of the node ID.

► Every node sees the network as a group of subtreesEvery node sees the network as a group of subtrees► For every subtree a node has a k-bucket in which For every subtree a node has a k-bucket in which

there is information about k nodes.there is information about k nodes.

SUBTREE MODEL K-BUCKETS (k=2)

Page 11: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

Routing modesRouting modes

SYMMETRIC RECURSIVE ITERATIVE

Page 12: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

OverSim - Overlay Network OverSim - Overlay Network SimulatorSimulator

► Based on OMNeT++ discrete event network Based on OMNeT++ discrete event network simulatorsimulator

► Consists of modules that are implemented in C++Consists of modules that are implemented in C++► In this study RELOAD was modeled in the DHT In this study RELOAD was modeled in the DHT

TestApp module.TestApp module.

Page 13: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

Input parametersInput parameters

► In the simulations 4 input parameters were In the simulations 4 input parameters were observedobserved

► Other parameters were kept fixedOther parameters were kept fixed► Altogether 57 simulation scenarios were runAltogether 57 simulation scenarios were run► Simulation time for each simulation run was 14 Simulation time for each simulation run was 14

daysdays

Page 14: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

Output parametersOutput parameters

► Overall bandwidth usageOverall bandwidth usage► Message overhead for maintenanceMessage overhead for maintenance

Only when using Chord. Kademlia does not use separate Only when using Chord. Kademlia does not use separate maintenance messagesmaintenance messages

► Number of lookup hopsNumber of lookup hops Length of the lookup pathLength of the lookup path

► Lookup message overheadLookup message overhead Amount of data transferred in the lookup processes in Amount of data transferred in the lookup processes in

bytesbytes

► Lookup delayLookup delay Delay between sending a lookup and receiving an answerDelay between sending a lookup and receiving an answer

► Key distributionKey distribution Mean number of keys stored per nodeMean number of keys stored per node

► Lookup success rateLookup success rate

Page 15: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

► Input paremeters affecting the mean total traffic are number of Input paremeters affecting the mean total traffic are number of nodes and key update intervalnodes and key update interval

► Kademlia uses more than twice as much bandwidth than ChordKademlia uses more than twice as much bandwidth than Chord► Iterative routing uses more bandwidth than recursive routing when Iterative routing uses more bandwidth than recursive routing when

Chord is usedChord is used► Mean total traffic is generally low (maximum bw for GPRS is ~14kB)Mean total traffic is generally low (maximum bw for GPRS is ~14kB)

ResultsResults

Page 16: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

ResultsResults► Kademlia needs less time to complete lookups than Chord Kademlia needs less time to complete lookups than Chord

doesdoes► Kademlia uses more bandwidth for lookup traffic than Chord Kademlia uses more bandwidth for lookup traffic than Chord

doesdoes

Page 17: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

► Kademlia gives higher lookup success rates than chordKademlia gives higher lookup success rates than chord► Recursive routing gives higher lookup success rates than iterative Recursive routing gives higher lookup success rates than iterative

routing when Chord is usedrouting when Chord is used► The shorter the key update interval the higher the lookup success rateThe shorter the key update interval the higher the lookup success rate► The longer the mean node lifetime the higher the lookup success rateThe longer the mean node lifetime the higher the lookup success rate

ResultsResults

Page 18: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

ConclusionsConclusions

► 2G mobile telephone networks can handle the 2G mobile telephone networks can handle the bandwidht usage of a P2PSIP network using bandwidht usage of a P2PSIP network using RELOAD peer protocolRELOAD peer protocol

► With the input parameters used in this study, With the input parameters used in this study, lookup success rates are in general too low for lookup success rates are in general too low for session establishment to work satisfyingly.session establishment to work satisfyingly.

► Kademlia outperforms Chord when lookup delay Kademlia outperforms Chord when lookup delay and success rates are comparedand success rates are compared

► Kademlia uses more than twice the bandwidth Kademlia uses more than twice the bandwidth that chord does.that chord does.

FUTURE RESEARCH:• This study could be developed by using more exact input parameters

• The reasons for the low lookup success rate could be investigated

Page 19: SIMULATING A MOBILE PEER-TO-PEER NETWORK Simo Sibakov Department of Communications and Networking (Comnet) Helsinki University of Technology Supervisor:

Questions?Questions?