Nomadic Sharing of Media: Proximity Delivery of Mass

Content within P2P Social Networks

Balázs Bakos

and Lóránt Farkas

Nokia Research Center, P.O. Box 392

H-1461 Budapest, Hungary

Abstract. In our paper we discuss a novel P2P file sharing system optimized

for smart phones, using proximity networking. Group management, peer dis-

covery and proximity multicasting are discussed. The experiment with the

proof-of-concept implementation on the Series S60 Symbian platform shows

that content sharing in social proximity can happen in a cost- and resource-

effective way.

1 Introduction

Today users can consume, create and share digital media content using their smart

phones, equipped with peer-to-peer (P2P) clients. P2P content sharing applications

have been very successful on the wired Internet. Napster [1], Gnutella [2] and BitTor-

rent [3] are among the most successful peer-to-peer (P2P) applications, measured to

be generating most of the Internet traffic today [5].

Search and network maintenance [6-12], speed and reliability [4,5], free-riding [16]

had been the main identified problems in P2P applications. Free-riding is becoming

increasingly acute now when P2P clients are increasingly turning mobile [14]. This

results in the decrease of the volume of available content, mobiles joining P2P over-

lay networks only for limited time periods, the overlay network being split into iso-

lated islands, hence the overall performance and reachability of the network degrad-

ing.

The community dimension of P2P networks evolved from community-friendly net-

works such as the first generation Napster, to anonymous ones because of the increas-

ing legal pressure, equivalent to the loss of the community dimension. Recently we

experience the reviving of the community dimension with the introduction of dark-

nets, small networks based on trust rather than anonymity. But trust brings the need

for these systems to be closed.

In our paper, instead of improving the existing P2P content sharing applications or

downscaling them to smart phones, we propose a novel P2P content sharing system

optimized for smart phones and proximity links. Section 2 describes the concept.

Section 3 focuses on social group management, enhanced peer discovery, multicast-

ing and user interface. Section 4 concludes this paper and outlines future directions to

be considered.

Bakos B. and Farkas L. (2006).

Nomadic Sharing of Media: Proximity Delivery of Mass Content within P2P Social Networks.

In Proceedings of the 3rd International Workshop on Ubiquitous Computing, pages 81-86

DOI: 10.5220/0002474500810086

 SciTePress

2 Nomadic Sharing of Media

The tragedy of the commons comes from the assumption that people consume rather

than share. In case of closed groups of mobiles we observe the opposite at least in two

different cases. First, people eagerly consume and share, if they think the content is

‘cool’ or of interest to other peers in the same closed group. Second, people eagerly

share content by others via forwarding, if this does not have as an effect a perceivable

degradation of the performance of their device.

In nomadic sharing of media content is shared by a user from his phone to a group of

people in the proximity. The process of content sharing consists of the following

operations:

− People share content with group(s). This means that the content (file(s) or

folder(s)) are selected and the group(s) are also selected.

− People meet friends. This is an opportunity for proximity scanning: the user’s

device scans and if devices belonging to the same group are found, content transfer

is initiated.

− Content is transferred to friends’ phones. Since multimedia content is typically

large, of several megabytes, and proximity events can be very short, it is possible

that in a session the content cannot be entirely transferred, which makes it neces-

sary to apply a segmentation mechanism. Namely, content is segmented into stand-

alone chunks of data.

− Friends’ phones carry shared content. The shared content can be the entire content

or chunks.

− Content (entire or chunks) reaches group members. If all chunks are available, then

the content is reassembled and consumed by the group members.

Nomadic share does fulfill the three criteria of darknets from [17] in a particular fash-

ion:

− the distributed content is not necessarily widely distributed, but expandable to such

a scenario. Currently there is only a closed group of users participating in the shar-

ing;

− users will not copy but share content which is interesting to themselves and think it

would be interesting for others as well (push rather than pull)

− the channels through which users are connected to each other is not high-

bandwidth in the strict sense: they are ad-hoc proximity links (of tens of kbyte/s).

3 Application Components

3.1 Social Group Management

One representation of our personal social network in the smart phone is the phone-

book or contact list. This is a good starting point for building a more complete data-

base. Since it is phone-centric in its structure (mobile phone number is in general a

default field of the phonebook records) it can be efficiently used as a starting point

for group creation using cellular short messages.

The contact list of smart phones provides the possibility to defining groups. However,

people are in general reluctant towards defining groups because of various reasons: it

is time consuming to define the groups, even if the groups are defined they are not

static and they can change during the time. Therefore selecting the proper social

group management method is a real dilemma for social interaction application de-

signers.

Several solutions can be envisioned to cope with this situation. First, the application

designer could rely on the fact that enthusiastic users will create the groups once they

have content to share. Ways of doing this represent invitations sent via short mes-

sages or invitations sent via proximity. The first scenario is relevant in the typical

case when the user sees or hears something and figures out enthusiastically how de-

lighted a group of his friends would be if they had the same experience. Sms is an

efficient way of quickly setting up a group and use it for content sharing. The second

case is encountered in cases when there is a social event and friends not having seen

each other since a long time want to share recent content such as weddings, birth of a

child etc. In this case setting up a group can be achieved by proximity messaging.

Second, there might be situations in which users are not very eager to make this step

by themselves. For instance, colleagues working in a team frequently meet at some

specific location and they call each other typically during daytime. In these cases the

smart phone can create groups in an automatic fashion or recommend associations

between individuals and groups. This is achieved by logging information related to

the user activity: the user’s communication patterns; the set of locations visited by the

user; the set of visible devices in the proximity/cellular surrounding of the user; the

contacts of the user, in the form in which the user stores them in the phonebook or

other possible repositories. Based on the collected data the phone can create clusters

and assign people to clusters using standard methods of statistics, directly or through

recommendations. Fig. 1 shows an exemplary event log containing 4 dimensions:

type of event, time, duration, originator, caller/callee.

Fr e dN/A> 8 hours8:00Proximity

Joh nN/A> 3 hours19:00Proximity

AliceOu t g o i n gN/A11:20SM S

BobOu t g o i n g21 min 20 sec10:53:21Ph o n e c a l l

Re mot e

part y

Direct i onDur at i onTi meType

Fr e dN/A> 8 hours8:00Proximity

Joh nN/A> 3 hours19:00Proximity

AliceOu t g o i n gN/A11:20SM S

BobOu t g o i n g21 min 20 sec10:53:21Ph o n e c a l l

Re mot e

part y

Direct i onDur at i onTi meType

09/03/2005

Fig. 1. Event log in four dimensions.

In nomadic share we followed a simple approach: the private group and the social

group concept were used. In the case of private groups, friends are selected and added

from the phonebook, via proximity scanning or by typing. The remote parties are not

explicitly aware of the existence of private groups: they exist as a formal group only

in the device of the group creator.

In the case of social groups members are invited and they accept or reject the invita-

tion. The remote parties are explicitly aware of their membership and can act upon,

by deciding whether to accept the invitation or not, and by choosing at any time in-

stant to leave the group or remain a member of the group.

3.2 Communications Engine – Peer Discovery and Multicasting in the

Proximity

In a typical use case users have some data files that they want to send to other mem-

bers of the closed social network, whom they occasionally meet. Unless the sender

and the target members are in proximity at the moment, the data transfer can take

place only when they meet next time. Such meetings are usually not planned but

happen spontaneously. The file sharing service is much more user friendly if it does

not require frequent user actions, so the user's phone needs to search for the target

phones on its own, by periodically running a discovery process in the background.

The discovery process typically uses the same proximity technology (e.g. Bluetooth)

as the data transfer. In order to increase the efficiency of the peer discovery, statistical

data collected about the location and the time of occurrence of past proximity events,

can be used. If the phone is equipped with a GPS, it can directly determine its loca-

tion. Otherwise the location is derived from the cell information of the cellular net-

work (list of cell ID: network ID doublets). To achieve sufficient accuracy, not only

the serving BS, but also the set of all visible BS-s can be monitored by the mobile

phone, the cell ID/net ID-s are collected and compared with stored cell ID/net ID

lists.

It is possible to store in the phone the most frequently met cell ID’s for each potential

peer, combined also with timing data. As a consequence, smart phones can predict

future proximity events and can concentrate their discovery effort to those locations

and occasions in which the targets occur with high probability. Therefore the prob-

ability of the proximity communications increases. On the other hand battery power is

saved this way, since the discovery process is executed more often only in cases when

there are good chances to find other peers in the proximity.

Content transfer in the proximity is performed via Bluetooth piconets, widely avail-

able in smart phones. In piconets up to 8 nodes can participate in data transmission,

but in a given time instant only two nodes communicate: a sender and a receiver. In

order to maximize the number of simultaneous communications taking place we em-

ployed parallel piconets consisting of 2 nodes each. Fig. 2 shows such a situation, for

the case of 8 nodes in each other’s proximity.

time

initial phase

efficient phase

f inal phase

Fig. 2. Efficient multicasting using small parallel piconets.

In Fig. 2. black nodes are sources (primary or secondary), grey nodes are also sources

but having received the content only partially. White nodes are recipients for the

content.

If the content is split to 2 chunks and marking by continuous arrow the transmission

of the first chunk and by dashed arrow the transmission of the second chunk, we can

split the transmission in 3 phases: initial, intermediate and final.

In the initial phase only few nodes are sources. In the intermediate phase there are as

many sources as recipients (or eventually one source less, the case of odd number of

nodes). In this phase all nodes (or eventually all nodes but one) are engaged in data

transmission, being either sources or destinations. In other terms piconets of 2 nodes

are performing in parallel in the intermediate phase. Therefore, if the Bluetooth inter-

ference is negligible, then the throughput in this phase is n/2 times higher than in the

one source – one destination case. Then in the final phase we have more potential

sources than destinations and the aggregate throughput decreases accordingly.

Fig. 2 does not give an algorithmic scheduling plan, it only shows one possible opti-

mal scheduling for 8 nodes. For scheduling plans the reader is referred to [13].

In the case when nodes are not in each other’s proximity, the principle remains the

same: the nodes store the data for themselves and also on behalf of other group mem-

bers, but the forwarding takes place later, when the other group members are seen.

Segmentation makes it possible to cope with the situation of nodes walking out of

coverage: the transmission can be resumed later without loss of data and the source

can be also another node, different from the one who started the data transmission

towards a particular recipient.

3.3 User Interface

Fig. 3 presents a set of snapshots from the application.

Fig. 3. Application snapshots: content info, group management, presence.

In the implementation phase the concept has been a standalone application with own

user interface. Snapshots display various views of this user interface: event notifica-

tion, group management, and presence are shown.

Notification is given to the user when content is received (in a persistent/non persis-

tent way). The user can check details of the received content, such as date and time of

creation, creator, sender.

Group management is available via creation/deletion of groups and adding/invitation

of members. The invitation can be accepted or rejected.

Presence info can be visualized: ‘green’ members have been encountered in the prox-

imity recently. Member details can also be visualized by clicking on the individual

members: address, phone number and group membership are displayed.

4 Conclusions

In our work we introduced a novel P2P file sharing application for smart phones to

deliver mass media content to a closed group of smart phones over multiple proximity

connections. The concept has been implemented on the Nokia 6630 platform. In a

user trial session enhancements were suggested to the concept related to the integra-

tion with existing Series S60 core applications such as sound and video recorder and

contact list. Such an integration would allow users to share content right at creation

time and thus minimize the number of steps required to initiate the content sharing.

Future work will focus on the merging of Internet-based P2P protocols and nomadic

sharing of media to exploit proximity technologies and context awareness available in

mobile devices. This direction may lead to the next generation of P2P protocols and

applications suitable both for personal computers and mobile devices.

References

1. The Napster protocol, http://opennap.sourceforge.net/napster.txt

2. The Gnutella protocol specification v0.4, document revision 1.2,

http://www9.limewire.com/developer/gnutella_protocol_0.4.pdf

3. Bram Cohen, Incentives Build Robustness in BitTorrent,

http://www.bittorrent.com/bittorrentecon.pdf, May 2003

4. Free riding on Gnutella is revisited: The Bell Tolls?,

http://csdl2.computer.org/comp/mags/ds/2005/06/o6001.pdf

5. Peer-to-peer file sharing – The impact of file sharing on service provider networks, white

paper, December 2000, http://www.sandvine.com

6. G. Csúcs, J.K. Nurminen, B. Bakos, L. Farkas, Peer-to-peer Protocol Evaluation in To-

pologies Resembling Wireless Networks. An Experiment with Gnutella Query Engine,

ICON’2003, Sydney, Australia

7. T. Hu, A. Sereviratne, General Clusters in Peer-to-Peer Networks, ICON’2003

8. K. Sripanidkulchai, B. Maggs, H. Zhang, Efficient Content Location Using Interest-Based

Locality in Peer-to-Peer Systems, Infocom’2003

9. H. Meling, A. Montresor, O. Babaoglu, Peer-to-peer Document Sharing using the Ant

Paradigm, NIK’2001, Tromso, Norway

10. E. Cohen, S. Shenker, Replication Strategies in Unstructured Peer-to-peer Networks, Sig-

Comm’2002, Pittsburgh, Pennsylvania

11. B. F. Cooper, H. Garcia-Molina, Modeling and Measuring Scalable Peer-to-peer Search

Networks, SIGCOMM’2002

12. B. F. Cooper, H. Garcia-Molina, Ad hoc, Self-supervising Peer-to-peer Search Networks,

Technical Report, Stanford University, 2003, www-db.stanford.edu/~cooperb/pubs

13. L. Farkas, B. Bakos, P. Spányi, A practical approach to multicasting in Bluetooth piconets,

accepted for publication in Proc. WCNC’2006, Las Vegas, April, 2006

14. http://symella.aut.bme.hu

15. J. Hummel, U. Lechner, Social profiles of virtual communities. In: Proc. of HICSS’2002,

Hawaii, 2002

16. J. Roettgers, Social networks: the future of p2p file sharing,

http://freebitflows.t0.or.at/f/about/roettgers

17. P. Biddle et al., The darknet and the future of content distribution, ACM Workshop on

digital rights management, Nov. 18, 2002, Washington DC, USA