Urban Scale Dissemination in Mobile Pervasive Computing

Environments

Alistair Morris, M

elanie Bouroche and Vinny Cahill

Distributed Systems Group, Trinity College Dublin, Dublin 2, Ireland

1 INTRODUCTION

The unprecedented urban growth (Capello, 2013) has

become a main issue characterising the 21st Century.

This growth has created new problems and exacer-

bated others. Urban sprawl, air pollution, trafﬁc bot-

tlenecks, insufﬁcient infrastructure form just small

subset of problems potentially remedied through the

collection of context data at an urban scale. This con-

text data enables the realisation of better decisions,

improved service delivery and greater accountability

through greater transparency.

To fully realise this concept and thus the po-

tential of Smart Cities the comprehension of facts

(Paskaleva, 2009) and thus context-awareness, de-

ﬁned as the ability to provide services with full aware-

ness of current execution environment, remains fun-

damental to the building of modern mobile and ubiq-

uitous systems. Several research efforts have pre-

viously devoted themselves to attempting to trans-

parently implement the features required by context-

aware applications including production, processing,

storage, and distribution of context data. This allows

the deployment of context-aware services in mobile

ubiquitous computing environments.

Context distribution, deﬁned as how we gather and

deliver relevant context about the environment to all

interested entities connected to the mobile ubiquitous

middleware, remains an unsolved problem and con-

tinues to emerge as a fundamental research area where

existing approaches do not scale. The presence of

non-negligible overhead hinders both the scalability

and reliability of a middleware (Corradi et al., 2010).

Detailing our arguments, this paper has the fol-

lowing structure: Initially we outline the Stage of re-

search and an Outline of objectives. Following this we

deﬁne the Research Problem and how we determine

this through conducting a review of Existing Work. Fi-

nally we propose the Methodology that we will apply

to test the Hypothesis and comment on what Future

Work remains.

2 STAGE OF RESEARCH

Research at present considers how we deliver context

data. Even if a range of solutions exist, it becomes

possible to compare the suitability of each particu-

lar deployment scenario. For instance, in an ad-hoc

network deployment, ﬂooding/gossip-based protocols

become more suitable where a lack of structured dis-

semination overlay ensures adaptability.

At present we remain focused in the process of ex-

tensively evaluating the main approaches that dissem-

inate context data. We hope to discover and deﬁne

the main beneﬁts and shortcomings of these. This be-

comes crucial when determining the scalability of any

context-aware application. Our principle research fo-

cus remains on how this affects context provisioning

to mobile pervasive devices in urban scale environ-

ments.

Differing from already existing work, our research

focuses on context data distribution to deeply study

the main requirements, implementation primitives,

and identify the main research challenges.

3 OUTLINE OF OBJECTIVES

A gap in research exists that hinders the realisation of

context-aware distribution in the wide-area wireless

networks that span the Smart City. We see that past

research has mainly focused on small-scale deploy-

ments, often limited to homes and commercial build-

ings, where context data distribution has affordable

run-time overhead. As a consequence, previous re-

search work has mainly addressed local middleware

issues in order to support context provisioning to sup-

ported applications. These approaches have also used

simple and centralised approaches to implement the

distribution process. Such approaches remain poorly

suited for the scalable middleware required in the

Smart City.

In addition, context data distribution involves dif-

ferent protocol layers (from OSI network to appli-

cation layer), and covers various emerging research

Morris A., Bouroche M. and Cahill V. (2014).

Urban Scale Dissemination in Mobile Pervasive Computing Environments.

In Doctoral Consortium, pages 18-22

 SCITEPRESS

ﬁelds, including traditional (infrastructure) mo-

bile ubiquitous systems, Mobile Ad-hoc NETworks

(MANET), Vehicular Ad-hoc NETworks (VANET),

and Delay Tolerant Networks (DTN) (Conti and Gior-

dano, 2007). As different network and middle-

ware deployments inﬂuence context provisioning ad-

ditional research remains a requirement as we do not

yet fully understand how they inﬂuence and limit the

scalability of context data distribution in the Smart

City.

4 RESEARCH PROBLEM

Context data distribution, namely the gathering and

delivery of relevant context data to all interested en-

tities connected in the mobile ubiquitous middleware,

has emerged as a new research area in context-aware

systems (Baldauf et al., 2007). In fact, context data

distribution remains signiﬁcant from both the per-

spective of the application and the middleware in the

Smart City. Service adaptation becomes triggered by

received context data: hence, the timely delivery of

context data enables services that promptly adapt to

the current execution context.

The middleware also has to transparently man-

age and route huge amounts of time sensitive context

data to mobile nodes: especially in wide-area mobile

networks, that can lead to non-negligible overhead,

thus hindering both system scalability and reliabil-

ity. Seminal surveys already argue the signiﬁcance

of context data distribution in context-aware middle-

ware. (Baldauf et al., 2007) (Sinderen et al., 2006).

5 EXISTING WORK

Dissemination enables data ﬂow from the source to

the sinks of the context data. Dissemination solutions

belong to three different categories: ﬂooding-based,

overlay-based, and gossip-based.

The ﬁrst two categories form deterministic ap-

proaches where a sink receives matching context data

produced by sources belonging to the same context

data distribution method. The last category takes the

form of a probabilistic approach meaning that a sink

has a likelihood to miss some matching data. Systems

adopting a hybrid approach that mixes these three

main dissemination solutions also exist. We present

examples of the use of these context data dissemina-

tion methods by citing well known systems.

5.1 Flooding-based

Flooding-based algorithms realise context data dis-

semination via ﬂooding operations. Here each node

broadcasts known context data across the entire net-

work. This means that receiver nodes must locally se-

lect context data as required. For instance, Adaptive

Trafﬁc Lights exchanges those context data useful to

co-ordinate red/yellow/green timings with vehicles at

an intersection (Gradinescu et al., 2007). HiBOp and

CAR ﬂoods neighbours with data useful for the main-

tenance of the DTN routing infrastructure (Boldrini et

al., 2007) (Musolesi and Mascolo, 2009). In MANIP,

each data comes with a physical locality tag that lim-

its the physical locality where data becomes ﬂooded

(Macedo et al., 2009).

5.2 Overlay-based

Overlay-based algorithms build dissemination “back-

bones” in a deterministic manner by using context

data subscriptions. This causes all data dissemination

to take place over a networked overlay.

To build the overlay nodes must exchange con-

trol information and this introduces overhead through

the additional communication that must take place.

Overlay-based approaches offer two visibility scopes

to each subscription for scalability: system wide

scope and limited scope. In network wide scope, the

dissemination process ensures that each subscription

remains visible to the whole network causing the re-

trieval of matching data to always remains possible.

With limited scope, the dissemination process limits

subscription visibility to a subset of nodes; unfortu-

nately this means that there exists a possibility that

the method will not ﬁnd the entire set of matching

context data.

The COPAL middleware employs centralised bro-

kers in which all subscriptions have network-wide

visibility scopes (Li et al., 2010). CMF similarly

uses distributed brokers that coordinate to supply

data to mobile nodes according to speciﬁc require-

ments (Kranenburg et al., 2006). Pervaho uses a

Location-based modiﬁcation of the Publish/Subscribe

(LPSS) paradigm by imposing location-based con-

straints: each publication and each subscription has a

visibility scope, and the delivery of published context

data to an active subscription only occurs if publisher

and subscriber lie in the intersection of these two

scopes (Eugster et al., 2009). Thus, Pervaho performs

context-/location-based ﬁltering to limit the amount

of events received. Here the adopted network layer

ensures that subscriptions have network wide visibil-

ity scopes. Finally, MobiSoC and MiddleWhere use

UrbanScaleDisseminationinMobilePervasiveComputingEnvironments

overlay-based methods to disseminate context data to

mobile nodes; at the same time, central servers pro-

vide network-wide visibility to context subscriptions

(Grupta et al., 2009).

We see a limited scope approach in EgoSpaces

that introduces a overlay-based approach based on

agents (Julien and Roman, 2006). Each agent oper-

ates over multiple views that includes context data

associated with hosts/agents in the physical local-

ity. Each view imposes constraints on contained

data/resources meta-data as each view spans close

neighbours. This leads to a limited visibility scope ap-

proach. Habit realises data dissemination in a limited

scope overlay-based manner by exploiting the phys-

ical proximity of nodes and the social relationships

of users (Mashhadi et al., 2009). It uses a regular-

ity graph that keeps trace of when and how often two

nodes communicate, and an interest graph that keeps

trace of nodes interests, to build dissemination paths.

Mobile Gaia groups nodes into clusters, and it uses

an event service to disseminate context data: because

every cluster has its own event service, the ﬁnal ap-

proach remains overlay-based with limited visibility

scope (Chetan et al., 2005).

5.3 Gossip-based

Gossip-based algorithms disseminate data in a prob-

abilistic manner by requiring each node to resend the

context data to a random set of neighbours. These

approaches suit fast-changing and unstable networks,

such as MANETs (Friedman et al., 2007) as they do

not attempt to construct and maintain complex rout-

ing infrastructure. Gossip-based protocols fall into

two board categories: context-oblivious and context-

aware approaches (Friedman et al., 2009).

Context-oblivious protocols use random re-

transmission probabilities. This means that no ex-

ternal context information becomes used to tailor

their operation. These include pure probabilistic gos-

sip techniques that simply resend each received data

with a re-transmission probability that differs for each

neighbour node and depends either on the local node

density or on neighbourhood information. No heavy

computation becomes necessary on traversed nodes as

the protocol selects nodes randomly. Unfortunately

this does not guarantee performance or scalability;

such protocols waste networking resources through

the gossiping of data unnecessarily.

Context-aware protocols select neighbouring

nodes for gossiping context data by using some

form of external context data. For instance, some

approaches use node distance to position context

data replicas; other approaches use social simi-

larity, for example the membership of a group, to

select the nodes for gossiping data to. Put simply,

context-aware approaches minimises the amount of

uselessly gossiped context data but they require more

co-ordination to exchange the context data used to

make gossiping decisions. For example, HiBOp and

CAR exploit context-awareness for message routing

purpose: above all, to select the best forwarder during

message routing (Boldrini et al., 2007) (Musolesi and

Mascolo, 2009).

5.4 Hybrid-based

Finally, hybrid approaches have become popular. Ac-

tive Highways collects context data ﬂooded from sen-

sors local/remote to the vehicles, and relies upon ﬁxed

servers to assist travelling vehicles; a overlay-based

approach with limited visibility scopes disseminates

data to servers (Iftode et al., 2008). Similarly, HiCon

exploits direct sensor access to produce data, and a

overlay-based approach to disseminate context data

by imposing locality scope (Cho et al., 2008). Finally,

in MobEyes, data becomes acquired either through

deployed sensor hardware or by other vehicles dur-

ing contacts: vehicles in proximity exchange data by

using a ﬂooding-based approach where each vehicle

broadcasts context data either locally sensed or re-

ceived by other vehicles according to the system con-

ﬁguration (Lee et al., 2006).

5.5 Conclusive Overview

Flooding-based and gossip-based methods do not

present good candidates for our pervasive Smart City

scenario: it appears neither can achieve both scalabil-

ity and reliability for context data delivery. Overlay-

based methods present an appealing approach but in a

pervasive mobile heterogeneous environment it lacks

feasibility where we must always rebuild dissemi-

nation backbones as the network exhibits mobility.

Hence we look closer at the hybrid based approach

where ﬂooding/overlay-based context data dissemi-

nation methods enable the delivery of control mes-

sages. With this an urban-scalable overlay-based con-

text data approach becomes possible.

6 METHODOLOGY

The NS3 network simulator will enable the examina-

tion of performance for each of the hybrid context

data dissemination protocols. NS-3 has become re-

graded as an efﬁcient and scalable simulator (Chau-

dray et al., 2012). This makes it possible to model

PECCS2014-DoctoralConsortium

the large heterogeneous networks encountered in the

Smart City. Therefore we aim to investigate perfor-

mance using mobility models deployed at the IP layer

with nodes densities that vary.

The simulation of mobility at the IP layer infers

the non-availability of end-points and the variation of

latency. We concentrate on two performance metrics

in particular Delivery Ratio and Dissemination Over-

head. This enables us to determine how mobility af-

fects performance and thus scalability.

We deﬁne respectively the Delivery Ratio as the

received context data packets divided by context

packets sent by the application and Dissemination

Overhead as the fraction of context data used by the

protocol for the dissemination of control messages.

Similar research efforts have previously employed

Delivery Ratio as a benchmark for efﬁciency (Tsiri-

gos and Haas, 2004) (Peng and Lu, 2000). Similarly

previous work also shows Dissemination Overhead

can provide an effect indicator (Dubois-Ferriere et al.,

2004) (Zhao et al., 2006).

To model a realistic Smart City scenario node den-

sities should exceed 10,000 nodes with the aim to sim-

ulate a context data dissemination network that has a

size comparable to 1% of the population of city of 1

million citizens. This will test the scalability and per-

formance of the context data dissemination method

and, importantly, the simulator itself.

7 HYPOTHESIS

We hypothesise that hybrid solutions that use differ-

ent dissemination algorithms together ensures better

performance and therefore scalability. For instance,

when overlay-based and ﬂooding/gossip-based meth-

ods become used at the same time: while the

overlay-based approach ensures context access, ﬂood-

ing/gossip based approaches can disseminate data in

a probabilistic manner thus reducing context access

time and increasing reliability.

Overall ﬂooding-based and gossip-based algo-

rithms appear promising in their performance and

thus their scalability. Even if ﬂooding-based meth-

ods have scalability issues, ﬂooding remains scalable

if constrained by scope.

Gossip-based approaches improve scalability by

reducing the delivery guarantees. In contrast to

context-oblivious approaches that waste networking

resources, context-aware gossip-based protocols rep-

resent a more efﬁcient way of building a context

data dissemination overlay. For example, HiBOp and

Habit uses social relationships as good hints to drive

gossip decisions (Boldrini et al., 2007) (Mashhadi et

al., 2009). Also, CAR demonstrates that the utili-

sation of low-level time context information, contact

frequency in particular, produces viable results (Mu-

solesi and Mascolo, 2009).

8 FUTURE WORK

Toward the main goal of realising how middleware

might adapt different context data dissemination algo-

rithms at run-time to foster greater performance and

urban scalability, additional research remains planned

that aims to deﬁne attributes that:

• Drives the selection of the proper context data dis-

semination algorithms at the run-time

• Enable the adaption of the run-time behaviour

for a speciﬁc context dissemination algorithm to

maximise performance and thus scalability

These attributes represent contextual modelled

and processed aspects useful for achieving scalabil-

ity. For example, (Taherkordi et al., 2008), models

the context attributes of a wireless sensor network that

then prompts adaption.

No solution adapts or switches different context

data dissemination algorithms at run-time to max-

imise scalability depending on the current status. Re-

search therefore remains to ensure that the adaption of

dissemination algorithms depending on run-time con-

ditions has the potential to enable urban scalable mo-

bile ubiquitous context data dissemination.

At the time of writing a thorough analysis of ex-

isting work has been conducted. The design and im-

plementation of the network simulations remains an

ongoing progress. The main challenge remains in

how simulations can become scalable while still ac-

curately reﬂecting the complex nature of the mobile

ubiquitous networks in the Smart City. To conclude

research also remains in how the speciﬁc implemen-

tation of a hybrid method may inﬂuence performance

and thus scalability. For example the notion of scope

in ﬂooding-overlay based hybrids or the choice of

context that drives context-aware gossip-overlay hy-

brids.

REFERENCES

Matthias Baldauf, Schahram Dustdar, and Florian Rosen-

berg. A survey on context-aware systems. Interna-

tional Journal of Ad Hoc and Ubiquitous Computing,

2(4):263–277, 2007.

Chiara Boldrini, Marco Conti, Jacopo Jacopini, and An-

drea Passarella. Hibop: a history based routing proto-

col for opportunistic networks. In World of Wireless,

UrbanScaleDisseminationinMobilePervasiveComputingEnvironments

Mobile and Multimedia Networks, 2007. WoWMoM

2007. IEEE International Symposium on a, pages 1–

12. IEEE, 2007.

Roberta Capello. Recent theoretical paradigms in urban

growth. European Planning Studies, 21(3):316–333,

2013.

Rachna Chaudhary, Shweta Sethi, Rita Keshari, and Sakshi

Goel. A study of comparison of network simulator-3

and network simulator-2. IJCSIT) International Jour-

nal of Computer Science and Information Technolo-

gies, 3(1):3085–3092, 2012.

Shiva Chetan, Jalal Al-Muhtadi, Roy Campbell, and

M Dennis Mickunas. Mobile gaia: a middleware for

ad-hoc pervasive computing. In Consumer Commu-

nications and Networking Conference, 2005. CCNC.

2005 Second IEEE, pages 223–228. IEEE, 2005.

Kyungmin Cho, Inseok Hwang, Seungwoo Kang, By-

oungjip Kim, Jinwon Lee, SangJeong Lee, Souneil

Park, Junehwa Song, and Yunseok Rhee. Hicon: a hi-

erarchical context monitoring and composition frame-

work for next-generation context-aware services. Net-

work, IEEE, 22(4):34–42, 2008.

Marco Conti and Silvia Giordano. Multihop ad hoc net-

working: The reality. Communications Magazine,

IEEE, 45(4):88–95, 2007.

A. Corradi, M. Fanelli, and L. Foschini. Adaptive con-

text data distribution with guaranteed quality for mo-

bile environments. In Wireless Pervasive Computing

(ISWPC), 2010 5th IEEE International Symposium

on, pages 373–380, 2010.

Henri Dubois-Ferriere, Deborah Estrin, and Thanos

Stathopoulos. Efﬁcient and practical query scoping in

sensor networks. In Mobile Ad-hoc and Sensor Sys-

tems, 2004 IEEE International Conference on, pages

564–566. IEEE, 2004.

Patrick Eugster, Beno

ıt Garbinato, and Adrian Holzer. Per-

vaho: A specialized middleware for mobile context-

aware applications. Electronic Commerce Research,

9(4):245–268, 2009.

Roy Friedman, Daniela Gavidia, Luis Rodrigues,

Aline Carneiro Viana, and Spyros Voulgaris.

Gossiping on manets: the beauty and the beast. ACM

SIGOPS Operating Systems Review, 41(5):67–74,

2007.

Roy Friedman, Anne-Marie Kermarrec, Hugo Miranda, and

ıs Rodrigues. Gossip-based dissemination. In Mid-

dleware for Network Eccentric and Mobile Applica-

tions, pages 169–190. Springer, 2009.

Victor Gradinescu, Cristian Gorgorin, Raluca Diaconescu,

Valentin Cristea, and Liviu Iftode. Adaptive trafﬁc

lights using car-to-car communication. In Vehicu-

lar Technology Conference, 2007. VTC2007-Spring.

IEEE 65th, pages 21–25. IEEE, 2007.

Ankur Gupta, Achir Kalra, Daniel Boston, and Cristian

Borcea. Mobisoc: a middleware for mobile social

computing applications. Mobile Networks and Appli-

cations, 14(1):35–52, 2009.

Liviu Iftode, Stephen Smaldone, Mario Gerla, and James

Misener. Active highways (position paper). In Per-

sonal, Indoor and Mobile Radio Communications,

2008. PIMRC 2008. IEEE 19th International Sympo-

sium on, pages 1–5. IEEE, 2008.

Christine Julien and G-C Roman. Egospaces: Facilitating

rapid development of context-aware mobile applica-

tions. Software Engineering, IEEE Transactions on,

32(5):281–298, 2006.

Uichin Lee, Biao Zhou, Mario Gerla, Eugenio Magistretti,

Paolo Bellavista, and Antonio Corradi. Mobeyes:

smart mobs for urban monitoring with a vehicular

sensor network. Wireless Communications, IEEE,

13(5):52–57, 2006.

Fei Li, Sanjin Sehic, and Schahram Dustdar. Copal: An

adaptive approach to context provisioning. In Wireless

and Mobile Computing, Networking and Communica-

tions (WiMob), 2010 IEEE 6th International Confer-

ence on, pages 286–293. IEEE, 2010.

D Macedo, A Dos Santos, Jos

e Marcos S Nogueira, and

Guy Pujolle. A distributed information repository for

autonomic context-aware manets. Network and Ser-

vice Management, IEEE Transactions on, 6(1):45–55,

2009.

Afra J Mashhadi, S Ben Mokhtar, and Licia Capra. Habit:

Leveraging human mobility and social network for

efﬁcient content dissemination in delay tolerant net-

works. In World of Wireless, Mobile and Multimedia

Networks & Workshops, 2009. WoWMoM 2009. IEEE

International Symposium on a, pages 1–6. IEEE,

2009.

Mirco Musolesi and Cecilia Mascolo. Car: context-aware

adaptive routing for delay-tolerant mobile networks.

Mobile Computing, IEEE Transactions on, 8(2):246–

260, 2009.

Krassimira Antonova Paskaleva. Enabling the smart city:

The progress of city e-governance in europe. Interna-

tional Journal of Innovation and Regional Develop-

ment, 1(4):405–422, 2009.

Wei Peng and Xi-Cheng Lu. On the reduction of broadcast

redundancy in mobile ad hoc networks. In Proceed-

ings of the 1st ACM international symposium on Mo-

bile ad hoc networking & computing, pages 129–130.

IEEE Press, 2000.

Amirhosein Taherkordi, Romain Rouvoy, Quan Le-Trung,

and Frank Eliassen. A self-adaptive context process-

ing framework for wireless sensor networks. In Pro-

ceedings of the 3rd international workshop on Middle-

ware for sensor networks, pages 7–12. ACM, 2008.

Aristotelis Tsirigos and Zygmunt J Haas. Analysis of mul-

tipath routing-part i: The effect on the packet delivery

ratio. Wireless Communications, IEEE Transactions

on, 3(1):138–146, 2004.

Herma Van Kranenburg, Mortaza S Bargh, Sorin Iacob,

and Arjan Peddemors. A context management frame-

work for supporting context-aware distributed appli-

cations. Communications Magazine, IEEE, 44(8):67–

74, 2006.

Marten J van Sinderen, Aart T van Halteren, Maarten Weg-

dam, Hendrik B Meeuwissen, and E Henk Eertink.

Supporting context-aware mobile applications: an in-

frastructure approach. Communications Magazine,

IEEE, 44(9):96–104, 2006.

Youping Zhao, Jeffrey H Reed, Shiwen Mao, and Kyung K

Bae. Overhead analysis for radio environment mapen-

abled cognitive radio networks. In Networking Tech-

nologies for Software Deﬁned Radio Networks, 2006.

SDR’06.1 st IEEE Workshop on, pages 18–25. IEEE,

2006.

PECCS2014-DoctoralConsortium