an architecture concept for mobile p2p file sharing services · high signalling traffic in...

4. Würzburger Workshop "IP Netzmanagement, IP Netzplanung und Optimierung"

An Architecture Concept forMobile P2P File Sharing Services

Andersen F.-U., Kappler C. (Siemens AG)Ivan Dedinski, Oberender J., de Meer H. (Universität Passau)

Tutschku K., Hossfeld T., Mäder A., Tran-Gia P. (Universität Würzburg)

University of WürzburgDepartment of DistributedSystems, Prof. Dr.-Ing. P. Tran-Gia

University of PassauInstitute of Computer Networks and Computer, Prof.Dr. H. de Meer

SIEMENS AG I&CMobile Networks Research & Concepts, F.-U. AndersenDr. C. Kappler

4. Würzburger Workshop "IP Netzmanagement, IP Netzplanung und Optimierung", 28.7.2004 Seite 2

Motivation (Trends and Assumptions)

Trends

Billing models for GPRS/UMTS: Towards time based and flat rateRising or evolving amount of „user created mobile content“

- Photos / small Videos of any (rights-free) kind- (self-composed) Ring tones, free text files / Information

ISP network traffic is dominated by peer-to-peer file sharing applications- Telefonica reports up to 80% fixed network p2p traffic

Assumptions

DRM: very complex, inter-disciplinary issue (but ouf of scope)- Alternatives: DRM light / Creative Commons Licence model

P2P experience is highly attractive to many users- Curiosity factor (what do other users have to offer?)


Objectives of MoPi („Mobile P2P“)

Problems for P2P in 2.5/3G mobile networksP2P traffic should be kept local for fixed/wireless operators

Asymmetric speed / P2P-unattractive charging

Variable terminal connectivity

High signalling traffic in unmodified P2P systems

Operators wish to participate in service creationQualitative + quantitative analysis of P2P in 2.5G / 3G

Approach / options:Operator-managed P2P services:Operator controls service and traffic dynamicsImproved P2P performance for mobile environmentsRealize in 3G by hybrid P2P structure

Introduce infrastructure elements to addressthe above problems

0 5 10 15 20 250

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

download time [h]

CD

F

with cachewithout cache


State of the art / related work

Smart cache for KaZaA (Joltid Inc.) and other P2P systemsman-in-the-middle approach. Protocol messages are intercepted and then tried to satisfy from a cache database

Project JXTA for mobile devices (J2ME, JXME)mobile P2P messaging / lightweight JXTA APIs for mobilesrelays enable connection to JXTA networksrendezvous peers with full capabilities (message routing / proxy services)no consideration of cellular mobile requirements

“A platform and applications for mobile peer-to-peer communications” (Kato et al., WWW2003, DoCoMo / Ericsson)

mobiles cannot run “full” P2P client software proxymixed architecture: hybrid, superpeer + pure P2P part.multicast overlay carries out multicast forwarding by multi-hop or multidestination unicast (ALM).


Option Space Cartography (I)

resource access

reso

urce

med

iatio

nce

ntra

lized

dece

ntra

lized

centralized decentralized

Proxy

MBMS

PureP2P

client/server

Classical operator domain

Classical user centric domain

Cache

Hybrid domain

IMS

e-donkey

VPN staticoverlay

resource access

reso

urce

med

iatio

nce

ntra

lized

dece

ntra

lized


Proxy

MBMS

PureP2P

client/server



Cache

Hybrid domain

IMS

e-donkey

VPN staticoverlay


Option Space Cartography (II)Shifting a pure P2P system towards the client-server model

resource control

reso

urce

med

iatio

nce

ntra

lized

dece

ntra

lized


e-Donkey

PureP2P

client/server



Hybrid domain

Solutionwith

e-Donkey+ Cache+ Proxy

resource control

reso

urce

med

iatio

nce

ntra

lized

dece

ntra

lized


e-Donkey

PureP2P

client/server



Hybrid domain

Solutionwith

e-Donkey+ Cache+ Proxy


Scope of work

Mobile P2P (MoPi)

Cellular mobileEnvironment

(2.5/3G)

Bluetooth, IR,WLAN, satellite, …

User servicesfor sharing

content

Network + controlservices, resilience

Hybrid P2P Pure P2P

Decisionstaken duringprojectdefinitionphase

Decisionstaken duringWP1 projectphase

Initial basic idea

Edonkey +derivatives JXTA Gnutella II,

Fasttrack, WinMX,

BitTorrentMobile-to-

Mobile P2POptimization ofPC-based P2Paccess via 3G

Fix-Mobile P2PInterconnect

DoCoMo-Ericssonproposal

4GLegend:

selected option

evaluated but notselected option


Design decisions

MBMS

Infrastructure options

Transfersession

IMS P2Papplication

server

Legend:

selected option

evaluated but notselected option

current work focus

P2PCache

EdonkeyIndex server

Mod

Algorithm options

Cachingstrategies

Presenceinformation

P2PCrawler

Option space

Distributionat GGSN

level

IMS / SIP Advancedsignalling

Proxy

SIPpresence

Superpeerorganization

Searchsession Mobile P2P

Architecture / Systemproposal


Caching Peer (pat. pend.) sub-system

P2P file exchange

Common P2P / overlay signaling

Indexserver

operator domain

Cachingpeer

Mob.peer

GGSN

# of downloads

triggerlevelfor cache

Air

traffi

c ge

nera

ted

Mob.peer

- Selected for final architecture; strong improvement of performance; imple-mentation without changes in client peer software

Caching peer specific signaling

Fixed P2Pnetwork

user domain


Crawling Peer (pat. pend.) sub-system

P2P file exchangeCommon P2P / overlay signaling

Caching peer specific signaling

- Selected for final architecture; allows for integration of mobile peers into the global P2P community

GGSN

Mob.peer

Mob.peer

Indexserver

operator domain

Crawlingpeer

user domain

Fixed P2Pnetwork


Smart Proxy (pat. pend.) sub-system

Mob.peer

Mob.peer

P2Pproxy

GGSN

Fire-wall

-Some of its core functionality can be taken over by the caching and crawling peers

- Deselected for final architecture; has lots of interesting options for further research (e.g. queue sharing)

P2P file exchangeLegacy mobile signaling

Common P2P / overlay signalingEnhanced MoPi signaling

Index-serverFixed

P2Pnetwork

operator domainuser domain


Proposed Architecture

CrawlingPeer

Cache Peer

0

2.5/3Gmobile network

operator domain

Mobile Peers

P2P file exchangemediation signalingenhanced P2P signalling

Mobile Control Domain (Presence Information)Internet Index Server

Internet Peers

user domain


Simulation model and first results

0 5 10 15 20 250

500

1000

1500

2000

2500

3000

Time [h]

Req

uest

ed fi

le

0 5 10 15 20 250

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

download time [h]

CD

F

with cachewithout cache

0 100 200 300 400 5000

0.5

1

1.5

2

2.5

3

3.5 x 104

time [min]

Tran

sfer

red

data

vol

ume

[kB

]

downlinkuplink

Popular file(s)

Mean value without cache: 11.2915 hMean value with cache: 5.8751 h

Simulation model type: Queue-based,event-oriented (virtual time base) and implemented in Java. Event inter-arrival time is randomized.


First results and conclusions(project is still running)

The project shows thatIt is possible to use P2P with GPRS with limitations The limitations can be addressed by using the described infrastructure elements / architecture

If the proposed architecture is used, it has the following advantages:

More performant P2P data exchange via the 2.5/3G cellular air IFIncreasing reliability of mobile-to-mobile file transfersCaching ( only 1 air interface needed)Based upon a fixed network P2P standard (eDonkey)Connecting to fixed network P2P world in a controlled wayOperator-provided index-server manages content mediationRedirection of clients to the operator‘s index server reducing inter-domain trafficFor the most parts, also applicable to fixed networksMotivation for end users to subscribe to a mobile P2P service because of a noticeable performance improvement

an architecture concept for mobile p2p file sharing services · high signalling traffic in...

Documents