ADS-Dir

ectory Services for Mobile Ad-Hoc Networks

Based on an Information Market Model

Christian Hutter, Matthias Brust and Steffen Rothkugel

University of Luxembourg

Department of Sciences, Technology and Communication

L-1359 Luxembourg, Luxembourg

Abstract. Ubiquitous computing based on small mobile devices using wireless

communication links is becoming very attractive. The computational power and

storage capacities provided allow the execution of sophisticated applications. Due

to the fact that sharing of information is a central problem for distributed appli-

cations, the development of self organizing middleware services providing high

level interfaces for information managing is essential. ADS is a directory service

for mobile ad-hoc networks dealing with local and nearby information as well as

providing access to distant information. The approach discussed throughout this

paper is based upon the concept of information markets.

1 Introduction

Today there are a large number of mobile devices like phones, personal digital assis-

tants and tablet PCs available. Most of them are equipped with wireless communication

adapters for local area communications. Storage capabilities and computational power

are continuously increasing, allowing users to run even sophisticated applications, not

only from the traditional domains like personal information management or business

applications but also offering entertainment, fun and recreation.

In this regard, using ad-hoc networks as underlying topology is promising due to

fostering ubiquity. Mobile ad-hoc networks are formed by large numbers of portable

devices which might not have a ﬁxed position and might join or leave the network

at any time. Low bandwidth, unreliability of wireless links, small transmission ranges

and unpredictable network topology changes pose technically challenging problems.

In realistic ad-hoc networks, resources and services need to be spread across a bigger

area than just the transmission range of a single device. Hence, devices are required

to communicate with one another by routing information through intermediate nodes,

resulting in so-called multi-hop ad-hoc networks. Aside from that, nodes should also

be considered as information providers. Users are supposed to collaborate with others,

introducing the need to share and disseminate information across multiple nodes.

This

research is supported by the Luxembourg Ministre de la Culture, de l’Enseignement Su-

prieur et de la Recherche. Any opinions, ﬁndings and conclusions or recommendations ex-

pressed in this paper are those of the authors and do not necessarily reﬂect the views of the

Luxembourg Ministre de la Culture, de l’Enseignement Suprieur et de la Recherche

Hutter C., Brust M. and Rothkugel S. (2004).

ADS-Directory Services for Mobile Ad-Hoc Networks Based on an Information Market Model.

In Proceedings of the 1st International Workshop on Ubiquitous Computing, pages 195-203

DOI: 10.5220/0002672701950203

 SciTePress

This paper introduces ADS, a generic middleware Ad-hoc Directory Service, aimed

at giving applications access to information located on more distant remote devices. The

subsequent sections are organized as follows. Section 2 discusses several distinct ad-hoc

application scenarios, in particular auction systems, ubiquitous multi-player gaming,

and M-Learning, motivating the usefulness of ADS. In section 3 a detailed description

of main ADS components is given. Section 4 compares ADS to similar middleware

solutions for mobile ad-hoc networks. Finally, section 5 gives a summary and provides

ideas for future work.

2 Application Scenarios

In order to illustrate the usefulness of ADS, several application scenarios in mobile

ad-hoc networks with different characteristics were presented in [1]. Firstly, UbiBay is

described, an auction system that can be considered as a counterpart to eBay for mobile

ad-hoc networks. Secondly, a learning scenario, an ad-hoc forum, is introduced, giving

an example of an educational application for mobile ad-hoc networks. Finally, gaming

applications are examined.

Auction System. UbiBay is an adaptation of prominent auction systems like eBay into

a mobile ad-hoc environment where users want to either sell items or bid on them.

Auctions are run by agents that populate marketplaces. A marketplace is a geographical

region where information is traded at given times. Users are not constrained to be at the

marketplace physically, but are allowed to utilize other ones mobile devices located at

the marketplace to let a software agent or a service installed on each device negotiate

with others on their behalf [2]. Applications query those marketplaces to get a list of

current auctions, allowing users to place bids.

M-Learning. In an M-Learning environment students are working collaboratively to

coordinate their effort. For instance before an exam, students should spend as much

time as possible for their preparations. Thus they want to use the time between two

lectures for studying. Unfortunately, students’ agendas typically are very individual,

making it extremely difﬁcult for them to form common study groups. To still be able to

join forces they might use an Ad-hoc Forum [3]. Students can use the according PDA

driven application to submit their questions to the forum. While passing or meeting

friends and fellow students, information like questions and answers are exchanged by

the PDAs. The system aims at ﬁnding solutions for the user’s questions or knowledge

lacks. Users are able to evaluate replies to the questions. The results of this procedure

are used to exclude non qualiﬁed answers.

Gaming. While traveling together, people can use their mobile devices for multi-player

gaming. For example a PDA might determine whether somebody nearby is interested in

joining a game. As soon as someone is found, the game integrates the new device into

the gaming community. All participating PDAs thereby form an ad-hoc network. This

scenario focuses mostly on groups of users located comparably close to each other like

in a train wagon or bus.

196

3 ADS: An Ad-Hoc Directory Service

ADS is an approach for a fully distributed directory service for mobile ad-hoc networks,

based upon an information market model. It allows the sharing of potentially replicated

information among applications running on the nodes of a mobile ad-hoc network. ADS

is supposed to be part of a middleware available on each participating device. Applica-

tions are supposed to be used in a cooperative way. The data those applications operate

on are initially generated on a particular device. This information might be useful for

the local device only, or might be shared with other instances of the application – re-

spectively with other users. In the latter case, the need for disseminating the information

in a controlled way arises.

3.1 Information Markets

Due to the very dynamic structure of the network and the potentially high number of

interacting devices, it is neither sensible nor possible to directly query any device. Fur-

thermore, in ad-hoc networks there is no notion of central servers. Hence, strategies for

collecting, exchanging, and gathering information are required. The approach proposed

in this paper for tackling this problem is to exploit existing characteristics of real-life

ad-hoc network infrastructure. In particular, it is possible and sensible to identify hot

spots, where the density of devices can be expected to be above a certain threshold.

Possible locations of such hot spots are central areas of cities, restaurants, cafeterias

or entrances of buildings like train stations, warehouses or universities. Because envi-

ronmental conditions might change during a day, however, locations of hot spots are

not stable. Thus the ADS needs to be context-aware, adapting itself to the current sit-

uation. The student cafeteria of a university for instance might be well visited during

daytime but deserted during nighttime or weekends. Based upon those hot spots, we

encourage their use for what we call information markets, concentrating and managing

large amounts of information in an adequate limited region and make it accessible to

interested applications. Initially, the ADS knows the location of at least one information

market. During runtime the platform discovers other markets, particularly their location

and available information. Either the moving devices themselves disseminate this infor-

mation in the network or it is propagated while communicating with an already known

information market. To determine their position devices use a positioning system, e.g.

GPS or a badge based system.

In order to make data available to applications, ADS supports two different types

of queries: synchronous local and asynchronous smart remote queries. Both types of

queries are useful by themselves, albeit a combination of them might be sensible to be

applied as well. Synchronous local queries are used to collect information from local

and nearby devices. They are called synchronous because they follow a synchronous

communication model. Asynchronous smart remote queries in turn enable applications

to retrieve data from information markets. They are asynchronous in nature, because

their execution time respectively behavior cannot be determined in advance.

197

Fig.1. Network scenario showing an example mobile ad-hoc network including an information

market

3.2 Synchronous Local Queries

The main purpose of synchronous local queries is to collect information that is avail-

able immediately. Firstly, the information maintained on the local device by the ADS

information broker component is retrieved. Optionally, the data might be augmented

by query results from nearby devices. This is realized by sending queries to the infor-

mation brokers in the immediate neighborhood. This information gathering process is

based upon a timeout mechanism to cope with the properties of ad-hoc networks prop-

erly.

Taking the example of gaming, synchronous local queries provide a suitable abstraction.

Interaction is restricted to the neighborhood. ADS might be used for instance to discover

currently active games together with information about players involved. Additionally,

slowly evolving game information e.g. related to trading can be handled appropriately

by ADS synchronous local queries.

In the M-Learning scenario, this type of query is well-suited to collect information for

instance while staying in the cafeteria. This way, knowledge from fellow students also

having a break can be retrieved directly. Nevertheless, in that case it is also sensible to

gather more information from other sources by referring to information markets. This

is realized by the second type of queries: asynchronous smart remote queries.

198

3.3 Asynchronous Smart Remote Queries

Aside from retrieving information from the immediate neighborhood, it shall be pos-

sible to consult information markets as well. Queries launched will be sent to the in-

formation market, starting to collect results there. Queries might stay active for a given

time, sending back results to the initiator in chunks. When transferring the query ADS

must use smart forwarding strategies, so that the query ﬁnally will arrive at the informa-

tion market. By tracking user movement it is possible to estimate the direction a device

is heading to in the near future.

Queries as well as responses can contain meta-data. Mandatory meta-data of queries

include information about the query initiator together with some contextual information

about his planned movements — e.g. taken from his calendar — that is used to deter-

mine time and location of where to send results to. Optional meta-data of queries might

for instance indicate an expected number of results. Knowledge about other information

markets can be propagated and collected by including it in the meta-data of responses.

Taking the example of UbiBay, asynchronous smart remote queries can be used

for example to gather information about ongoing auctions respectively items offered.

In the Ad-hoc Forum scenario, questions as well as answers are collected and pooled.

Hence, the number as well as the quality of the knowledge available at the market can

be expected to be high, especially compared to the cafeteria case aforementioned.

3.4 Information Market Management

Information markets are supposed to manage potentially large amounts of data. This is

done in a distributed and replicated way. The management strategies are described in

the subsequent sections.

The IMM. The devices composing an information market need to act in a self-organizing

way. When a device enters the information market, it contacts the information market

manager (IMM). If there is no response, a new IMM is created on that device. The IMM

collects the information about the free capacity available on the information market and

about replicated information on each device. In case multiple IMMs are created at the

same time, e.g. when several devices arrive simultaneously at an empty market, they

need to select a leader using known leader selection algorithms for mobile ad-hoc net-

works. The IMM assigns new information which arrives at the market to devices with

free capacities. Depending on the level of importance of the information it might be-

come replicated on several hosts. Devices to which the IMM already assigned data are

called active while devices not holding data are passive to the information market.

The IMM needs to use load balancing algorithms to distribute information across the

devices available, depending on their free capacities. As all information is added to the

system by the IMM it can keep track about the type of information which is hosted on

its market. This meta-information can be added to asynchronous smart remote queries

to improve the chances of retrieving relevant information fast.

199

Dealing with Mobility. When a passive device leaves the area of an information market

there is no loss of information. In case an active device A moves out of an information

market its data needs to be transferred back to the market. A transfers the data to another

device B which is still at the market or at least heading into the right direction. If B

is already within the area of the information market it forwards the data to the IMM

for redistribution. Otherwise it keeps the data until reaching the information market to

ﬁnally forward it. If the direction of B changes before it enters the information market,

B transfers the data to yet another device heading towards the market.

A special case occurs if the IMM leaves the information market. However, the same

strategy used by active devices when leaving the market is applied. Due to race condi-

tions, a new IMM might have been created in between. In that case, the old IMM will

deactivate itself. To increase the probability of the IMM staying inside the market it

tries to reside in its center. If needed, the role of the IMM is taken over by a different

device.

Fig.2. Zoom to the information market showing (moving) active and passive devices and an IMM

Communication within an Information Market. Information markets differ from the

rest of the ad-hoc network. Due to the high amount of available devices within a limited

geographic area it is possible to establish a fast communication and information distri-

bution between the participating devices. Even if they are not in direct communication

range it is possible to use other nodes to forward messages using ﬂooding techniques in

this limited area. Messages sent to the IMM only need to be forwarded to the center of

the market as the IMM is supposed to stay in this area.

200

Disaster Recovery for Active/Passive Market Devices. Disaster recovery needs to be

done when a device leaves a market without being able to perform the normal ”leave

market” actions mentioned above. This can happen as a result of a power-down or

hardware respectively software error. In the case the device was passive to the market no

information is lost. In the case of an active device, all information that is not replicated

is also lost. The IMM can always try to restore replicated information of devices which

are able of sending a ﬁnal sign off message before shutting down. If a device can not

send this message it will depend on the strategy chosen if the degree of replication can

be reestablished (cf. also the paragraph describing information replication below).

Disaster Recovery for the IMM. The loss of the IMM will be detected when infor-

mation will be added to the system or a new device enters the information market. At

that time, a new instance of the IMM will be created. During initialization of the new

IMM it asks all active or passive market devices about their free capacities. In case sev-

eral IMMs are created, leader election techniques are used to elect one of them (c.f. the

paragraph 3.4 about the IMM ). Basically, in the current design the IMM is a central

entity. However, in case of failures, it is possible to create a new IMM dynamically. The

state of the IMM can be reconstructed by gathering according data from the devices on

the market.

Publishing Information on an Information Market. When an application or service

wants to publish information, it needs to send it to an information market. The selection

of a target market can be done context-aware by several strategies, e.g. by aspects of

best ﬁtting in terms of types of information already managed by a market or by distance

to targeted applications. In both cases the information is forwarded to a market using the

same strategies as applied in the asynchronous smart remote queries. The IMM handles

the dissemination and replication of newly available data, taking into account meta-data

like lifetime or replication degree.

Information Replication. The ADS allows information replication on information

markets. If the IMM gets information of high importance, as indicated by meta-data

determining the replication degree, it will assign this information to multiple devices.

For reasons of recoverability those nodes need to know that they are now managing

replicated information. To prevent information loss the IMM also keeps a list of the

replicated information. The ADS provides two strategies to keep the number of copies

at the suggested value, differing in the bandwidth usage. The ﬁrst strategy requires

devices to be able to send a short sign-off message when leaving a market. When the

IMM receives such a message from devices hosting replicated information it starts a

recovery operation, i.e. it gets and assigns a new copy of the information to another

device. If a device is no longer available and was not able to send a sign-off message

one copy will be lost. To prevent this, in the second strategy the IMM periodically

checks if all devices involved into a replication process are still available. If necessary,

the aforementioned recovery operation is triggered.

201

4 Related Work

There are a number of well known directory services for traditional networks environ-

ments like OpenLDAP and Microsoft’s Active Directory. But due to the very dynamic

structure of mobile ad hoc environments together with the restrictions in terms of pro-

cessing power and storage capabilities, those are not suitable directly. ADS is explicitly

designed for use in such environments.

The application scenarios introduced require information to be stored for further

usage. A wide variety of different application-dependent information types needs to be

managed. Simple strategies like ﬂooding the network by disseminating data using adap-

tive protocols like introduced in [4] are not reasonable in general. A different approach

is used in NOM [5] which forwards queries and creates responses from every node in

the network. In both strategies, broadcast storms are likely to occur. SIDE Surfer [6] in

turn only allows the automatic exchange between directly connected devices based on

user proﬁles, giving the applications access to a limited set of information only. ADS

allows applications to use the synchronous local queries to access such information

and additionally provides asynchronous smart remote queries for retrieving data from

information pools currently available on information markets.

Both TOTA (”Tuples on the air” [7]) which aims at supporting adaptive context-

aware activities and MeshMdl [8] use the tuple paradigm in mobile ad hoc networks.

But they miss the concept of information replication which increases the probability of

loosing important information.

5 Conclusion and Future Work

ADS is a middleware service designed for use in mobile multi-hop ad-hoc environ-

ments. It enables applications to directly share information with as well as receive

information from the devices in the neighborhood, together with providing access to

information markets. The main idea of introducing information markets is to be able to

identify well-known places where different kinds of information from multiple appli-

cations can be pooled and exchanged. Within information markets, sophisticated algo-

rithms can be applied e.g. in terms of load balancing and fault tolerance by replication.

The interaction with information markets is facilitated by the asynchronous smart re-

mote queries, which are long-term queries. After being launched, these queries travel

to information markets, gathering information there to be sent back to their initiator.

This makes the information market model well-suited for data that remains stable for a

certain time.

The environment envisaged is neither stable nor fully predictable. The system might

for instance suffer from network partitioning, resulting e.g. in queries reaching infor-

mation markets either late or possibly not at all. Another problem occurs if the number

of participating devices drops under a certain threshold. Then, proper dissemination and

replication of information within information markets cannot be guaranteed anymore.

Strong guarantees cannot be given in such kind of environments anyway. It is only

possible to minimize the impact of different kinds of failures and shortcomings. In the

model proposed, this is done through the use of information markets together with their

202

management strategies. As shown in [2], the information market model is a promising

approach. Nevertheless it is required to do a more detailed evaluation of the proposed

concepts.

In the future, several interesting aspects need to be evaluated. For instance, the man-

agement of information about marketplaces themselves needs to be tackled, like their

establishment and discovery. One possible solution here is to rely on general knowl-

edge which locations might be well-suited as e.g. train stations and shopping malls. A

different approach is to dynamically determine marketplaces by observing devices and

discovering common patterns. The management of data within information markets is

another area for future research, including for example to de-centralize the Information

Market Manager. It might also be interesting to evaluate if an exchange of information

across several information markets would improve the overall system effectiveness. Ad-

ditionally, we currently assume information to be stored atomically. In order to leverage

both fault tolerance and load balancing, it could be interesting to split one piece of in-

formation into multiple chunks obeying a certain level of redundancy, and to distribute

those on several devices. Finally, several application domains exist that obey different

interaction paradigms. Examples include cases where data becomes out-of-date rather

quickly, or applications that are comparatively tightly coupled. Hence, other concepts

respectively communication patterns aside from the information market model for shar-

ing and exchanging information need to be developed.

References

1. Hogie, L., Hutter, C.: Middleware Support for Ubiquitous Computing. ERCIM News No. 54,

July (2003)

2. Goergen, D., Frey, H., Lehnert, J.K., Sturm, P.: Marketplaces as Communication Patterns in

Mobile Ad-Hoc Networks. Proceedings of KiVS 2003, Kommunikation in Verteilten Syste-

men (2003)

3. Brust, M., Rothkugel, S.: Cooperative and Mobile Learning in Ad-Hoc Networks. Proceedings

of WCETE 2004, World Conference on Engineering and Technology Education, Brasil (2004)

4. Kulik, J., Rabiner, W.,Balakrishnan, H.: Adaptive Protocols for Information Dissemination in

Wireless Sensor Networks. Proceedings of the 5th ACM/IEEE Mobicom Conference, Seattle,

WA, August (1999)

5. Doval, D.,O’Mahony, D.: Nom: Resource Location and Discovery for Ad Hoc Mobile Net-

works. Proceedings of the 1st Annual Mediterranean Ad Hoc Networking Workshop, Med-

hoc-Net (2002)

6. Touzet, D., Menaud, J.-M., Weis, F., Couderc, P., Bantre, M.: SIDE Surfer: A Spontaneous

Information Discovery and Exchange System. Proceedings of the 2nd International Workshop

on Ubiquitous Computing and Communication, Barcelona, Spain (2001)

7. Mamei, M., Zambonelli, F., Leonardi, L.: Tuples on the Air: A Middleware for Context-Aware

Pervasive Computing. Technical Report No. DISMI-2002-23, University of Modena and Reg-

gio Emilia, Italy, August (2002)

8. Herrmann, K.: MESHMdl - A Middleware for Self-Organization in Ad hoc Networks. Pro-

ceedings of the 1st International Workshop on Mobile Distributed Computing (MDC’03),

Providence, Rhode Island, USA (2003)

203