14
Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese 1,2 , Paulo Sim˜ oes 1 and Fernando Boavida 1 1 Centre for Informatics and Systems of the University of Coimbra - CISUC Department of Informatics Engineering - DEI University of Coimbra - UC - Portugal [email protected] 2 Departament of Computer Science - DCC University of the State of Santa Catarina - UDESC 889219-710 Joinville, SC, Brazil [email protected] Jun 18, 2012 Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 1 / 14

Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Embed Size (px)

Citation preview

Page 1: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Peer Selection in P2P Service Overlays usingGeographical Location Criteria

Adriano Fiorese1,2, Paulo Simoes1 and Fernando Boavida1

1Centre for Informatics and Systems of the University of Coimbra - CISUCDepartment of Informatics Engineering - DEI

University of Coimbra - UC - [email protected]

2Departament of Computer Science - DCCUniversity of the State of Santa Catarina - UDESC

889219-710 Joinville, SC, [email protected]

Jun 18, 2012Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 1 / 14

Page 2: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Outline

1 IntroductionBackground (BPSS)

2 ValidationBPSS Used MetricValidation AspectsSelected approachPreliminary Results

3 Conclusions

4 Future Work

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 2 / 14

Page 3: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Introduction

Services and service components are becoming basic elements ofinteraction among service providers.

Service Overlay Networks (SON) may act as an infrastructure whereservices are published/offered and to which users access in order tolocate, select and use those services.

P2P technology helps to create self-organizing Service OverlayNetworks in several levels.

Therefore, services and service components are offered in a P2P SON.

In order to maximize performance, the best peer must be found in theP2P SON, among all the potential partners that provide the desiredservice.

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 3 / 14

Page 4: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Best Peer Selection Service (BPSS)

BPSS is part of an architecture for Services Management in P2Penvironments called OMAN, proposed previously in Fiorese et al.,OMAN - A Management Architecture for P2P Service OverlayNetworks, AIMS 2010.

BPSS is responsible for select the best peer for interacting with arequesting peer, regarding a particular service and metric.

BPSS runs over a P2P SON at the same level of a second overlay-tiercalled AgS responsible for speedup search operations.

SON peers can request best peer information (select BP), regarding aparticular service, from the BPSS module.

BPSS answers with the best peer regarding the agreeded metric.

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 4 / 14

Page 5: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

OMAN Architecture

P2P Search

Service

Aggregation

Service

Support for

Legacy

Management

Systems

Application/Service

2nd

Layer

1st Layer

3rd

Layer

Configuration

Manager

Resilience

QoE

AutonomicBest Peer Selection Service

Overlay Monitoring

Support for P2P CommunicationSON

provisioning

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 5 / 14

Page 6: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Best Peer Selection Service (BPSS) Framework

Domain CDomain BDomain A

Provider/User

Aggregation Service (AgS)

SON peer

Aggregation peer

Aggregation links

Physical links

Overlay links

Query

SON

Best Peer

Selection Service

(BPSS)Query

Se

lect B

P

BP

List of BP

Select BP

BP

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 6 / 14

Page 7: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

BPSS Used Metric

Developed by Kaune et al. (2009) - Modelling the Internet delay spacebased on geographical locations.

It is based on a predictive model of the Internet delay space that takesinto account the geographical location of nodes and the delay betweenthem.

It uses a rich set of real Internet measured data namely end-to-endround trip time and end-to-end link jitter.

These data are mapped into a 5-dimensions Euclidean space model ofthe Internet combining it with global network positioning information.

Metric

Using the coordinates of each peer in this 5-dimensions model, it is thenpossible to calculate the Euclidean distance between peers, taking intoaccount not only network conditions but also peer location.

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 7 / 14

Page 8: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Validation Aspects

Five geographical domains comprising Portugal, Spain, France, Italyand Germany made part of the experiments.

All the peers requesting best peers were from in the Portuguese domain.

Comparison between the best and second best peer selected.

Comparisons were made by domain.

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 8 / 14

Page 9: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Validation (selected approach)

Evaluation by simulation.

Several sizes of P2P SON were simulated, ranging from 50 to 300 P2PSON peers, with steps of 25 peers.

Parameters:

Average of 10 repetitions of 100 best peer selection operations for eachP2P SON sizes.

Each execution simulated 50 hours of BPSS work.

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 9 / 14

Page 10: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Best Peer Selection Execution

Bars for each group correspond to Portugal, Spain, France , Italy, and Germany, respectively from the left to the right.

0

10

20

30

40

50

50 75 100 125 150 175 200 225 250 275 300

Num

ber

of B

est p

eers

Number of SON peersAdriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 10 / 14

Page 11: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Averaged geographical distribution of best peer selection

5

10

15

20

25

30

35

Portugal Spain France Italy Germany

Num

ber

of B

est p

eers

Domains

Best Peer Distribution on Average

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 11 / 14

Page 12: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Summing up

Taking the sum of best peers and second-best peers by domain, SONpeers in the requester’s domain (Portugal) were selected as best peersin 27% of the time, followed by Spain (22.5%), Italy (22%), Germany(14.5%) and France (13%).

This means that almost half the best peer selections resulted in peersbelonging to the same geographical domain or to the neighboringgeographical domain.

Obtained results have shown that BPSS performs well and that theoverall OMAN architecture - of which the AgS service is a keycomponent - is very effective.

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 12 / 14

Page 13: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

Future Work

Compare these results with the BPSS utilization of other selectionmetrics, including related with QoS.

Use of the OMAN approach to identify ways of maximizing user QoE.

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 13 / 14

Page 14: Peer Selection in P2P Service Overlays using Geographical Location Criteria Adriano Fiorese - State University of Santa Catarina /University of Coimbra Paulo Simões, Fernando Boavida

The End!

Thank you!

Questions/Comments/Suggestions/Discussion?

Contacts

Personal Web Page: (http://eden.dei.uc.pt/˜fiorese)

Laboratory of Communications and Telematics at UC(http://www.cisuc.uc.pt/lct/)([email protected])

GRADIS - Group of Network and Distributed Applications at UDESC(http://www2.joinville.udesc.br/˜gradis/)([email protected])

Adriano Fiorese (UC,UDESC) ICCSA 2012 Jun 18, 2012 14 / 14