Evaluation of Ad hoc Routing Protocols Evaluation of Ad hoc Routing Protocols under a Peer-to-Peer Applicationunder a Peer-to-Peer Application

Authors: Leonardo Barbosa

Isabela Siqueira

Antonio A. Loureiro

Federal University of Minas Gerais – Brazil

Computer Science Department - http://www.dcc.ufmg.br

SummarySummary

Introduction Motivation P2P implemented protocol Simulation and results Conclusion Questions

What is P2P?What is P2P?

Distributed paradigm Dynamic topology Nodes have equivalent funcionalities and

provision capacities (peers) Peers play the role of servers and clients

simultaneously (servents)

What are P2P applications?What are P2P applications?

P2P sharing systems Able to share

Hard disk spaceFilesCPU

Phenomena in the Internet Examples: Gnutella, Freenet, Napster, ICQ

P2P Applications and MANETsP2P Applications and MANETs

P2P paradigm is the basis for both MANETs and P2P applications Central units are non-existent in both

environments Their nodes are clients and servers at the

same time Self-organizing networks Responsible for routing queries in a

distributed environment

Faithfulness to the ModelFaithfulness to the Model

P2P applications Usually built over a

network based on the Client/Server model

Clients of foreign servers

Neighbors might be geographically many hops apart

MANETs: Implement their

own communication mechanism

Only communicate with servents

Peers are only a single-hop away from their neighbors

MotivationMotivation

Similarities between the systems Scarcity of work in which both systems coexist Could P2P applications become “killer

applications” in MANETs?

By means of this...By means of this...

Direction and Speed

Out of Range

Ad hoc Connection

Transmission Range

MANET

... and this ...... and this ...

RouterPeerConnection among P2P application nodesConnection among routers and peersConnection among routers

P2P Application Network

... this was built!... this was built!

Ad hoc node P2P application nodeConnection among P2P application nodesConnection among application and MANET nodes

P2P Application Network over a MANET

P2P Implemented ProtocolP2P Implemented Protocol

Based on Gnutella Joining the network

Transmission of a broadcast message searching for neighbors

BROADCAST-SEND Achieved peers respond

BROADCAST-REPLY Neighbors election

P2P Implemented ProtocolP2P Implemented Protocol

Searching Query transmissions to neighbor peers

QUERY-SEND Process goes on until the information is found or

dropped In case it is found, the servent that owns the file

wanted responds to the “query-source” peer QUERY-REPLY

Transferring files Establishment of an end-to-end communication Fragmentation and transference of the

information

P2P Implemented Protocol P2P Implemented Protocol

Controlled flooding Each peer has a cache to avoid a request

being handled twice P2P header includes TTL

Connectivity control PING and PONG messages

Simulation Simulation

40 mobile nodes, 12 executing an instance of the P2P application

Grid: 200m x 200m MAC protocol: IEEE 802.11 Energy consumption: 0.3W (Tx), 0.2W (Rx) Each scenario was simulated 33 times Simulation time: 300 s

Simulation Simulation

Evaluated protocols: Dynamic Source Routing Protocol (DSR) Destination-Sequenced Distance-Vector

Routing (DSDV) Ad Hoc On Demand Distance-Vector (AODV)

Analysis MethodsAnalysis Methods

Workload Amount of requests per peer Size of shared files

Mobility Pause time Speed

Network Density Number of nodes Transmission range

Amount of Peers

WorkloadWorkload

DSDV presented the most overhead and stability when the number of queries were augmented (2000 packets on average)

The behavior of the other protocols were similar

Queries (amount)

Overh

ead

(p

kts

)

MobilityMobility

Latency as function of speed increased exponentially when DSR protocol was used (climbed to 10x more) All protocols provided more information unavailability and worse P2P connectivity in low levels of mobility

Late

ncy (

s)

Speed (m/s)

Network DensityNetwork Density

The curves of the delivery ratio as function of the amount of nodes behaved equivalently for the three protocols

The curves climbed rapidly when the network was little densed Above 20 nodes delivery ratio was between 60 and 70%

Nodes (amount)

Delivery

R

ati

o

(%)

Peers (amount)

PeersPeers

H

op

s (

nu

mb

er)

DSR was the protocol that calculated the worst routes and had

the major impact when the number of peers increased There was an increase in the range from 10 to 20 peers

ConclusionConclusion

“There is no silver bullet”, each of the protocols analysed performed well in some scenarios for some metrics while had drawbacks in others

It is important to identify accurately characteristics of the P2P application (load, amount of peers etc.) in order to opt for a protocol

Questions? Questions?

Thank you for your attention! Contacts

E-mail: {leob,isabela,loureiro}@dcc.ufmg.brHome Page: http://www.dcc.ufmg.br/~leobResearch Group Site:http://www.lecom.dcc.ufmg.br/~tbb