See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/37686383

Design of a Backup Network for Catastrophe Scenarios

Conference Paper · June 2009

DOI: 10.1145/1582379.1582512 · Source: OAI

CITATIONS READS
4 250

4 authors, including:

Boris Koldehofe Hugo Miranda

Technische Universität Ilmenau University of Lisbon
161 PUBLICATIONS 2,755 CITATIONS 90 PUBLICATIONS 959 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Hugo Miranda on 27 May 2014.

The user has requested enhancement of the downloaded file.

 Design of a Backup Network for Catastrophe Scenarios
Position Paper
S. Alves B. Koldehofe
Portugal IPVS Stuttgart, Germany
H. Miranda F. Taiani
University of Lisbon, Portugal University of Lancaster, UK

ABSTRACT to rely on communication mechanisms that allow manage-

Communication networks play a fundamental role in the re- ment teams to: i) learn about the occurrence of new events
sponse to a massive catastrophe, like an earthquake or a as they unfold; ii) continuously gather feedback from the
large-scale terrorist attack to a major urban area. In such personnel on the ground; and iii) deliver orders or updates.
situations, command centres must be able to rely on a fully In an effort to ensure optimal communication, military
operational communication network, for example to learn and civil defence forces of most countries reserve part of the
about on-going situations and allocate and guide the rescue wireless radio spectrum for their networks and have a pri-
teams. Communication is bidirectional: once in the field, vate wireless communication infrastructure deployed. A lot
these teams will feed the command centre with a more ac- of research effort has been spend on setting up such spe-
curate view of the situation, contributing to the efficient cialised networks we refer to as rescue networks (e.g. [5,
allocation of the resources. Failures in this network, even if 13, 24, 11, 2]). While rescue networks already consider to
localised to some of the regions affected by the catastrophe, support (ad-hoc) communication over heterogeneous devices
can have costs both monetary and in human lives. and networks, they typically neglect interactions with de-
In this position paper, we propose the creation of a redun- vices provided by civilians in order to ensure best service
dant, best-effort, emergency communication network that performance for the rescue teams. However, the early de-
could serve to mitigate localised failures using off-the-shelf ployment of the rescue teams will be mostly dictated by
widespread technology. We give an overview of an architec- the events learnt by the coordination authority from the
ture for a backup network, highlight the possible advantage alerts provided by civilians. At this stage, the civil network,
of such an architecture to disaster management and discuss composed of the wired and wireless network infrastructure
challenges that need to be overcome in realising it. operated by telecommunication companies, will play a fun-
damental role, given that civilians do not have access to the
rescue network.
Categories and Subject Descriptors The infrastructure supporting civil and rescue networks
C.2.5 [Local and Wide-Area Networks]: Internet; C.2.2 should be considered as vulnerable to the catastrophe as any
[Network Protocols]: Protocol architecture other infrastructure. Falling antennas or interference from
malfunction devices can disrupt communication in some re-
1. INTRODUCTION gions. Disruption will especially affect the infrastructure of
civil networks, which is not designed to be as resilient as res-
In large urban areas, catastrophes like earthquakes, fires
cue networks.1 However, one should not exclude the possi-
or massive terrorist attacks are more subject to acquire large
bility of disruptions on the rescue network. Besides damages
proportion in both human and infra-structural costs. After
on constructions and on the hardware, terrorists could in-
such an event, public authorities are expected, among oth-
clude the rescue network infrastructure in their preferential
ers, to locate and rescue the victims; preserve or restore
targets to maximise the cost, in lives, of an attack.
public order and identify locations where threats persist or
In this paper, we discuss the construction of a Backup
are imminent. Disaster management is usually delegated to
Network to help mitigate the adverse impact of failures in
a central authority which coordinates all the teams deployed
both the civil and rescue networks. The goal is to provide
within the catastrophe area.
coverage in locations where Rescue and/or Civil Networks
In such a scenario, a reliable communication infrastructure
have been affected. The network is composed by off-the-
is fundamental to perform an optimal allocation of the avail-
shelf equipment commercially available and puts together a
able resources. To be effective, coordination must be able
number of technologies that have recently gained attention.
Components of this network are made available by civil-
ians and rescue teams, in response to the occurrence of the
Permission to make digital or hard copies of all or part of this work for catastrophe. The network provides a best effort service, un-
personal or classroom use is granted without fee provided that copies are der the perspective that some connectivity would be better
not made or distributed for profit or commercial advantage and that copies than none. In the line of what has been proposed for ad-hoc
bear this notice and the full citation on the first page. To copy otherwise, to
republish, to post on servers or to redistribute to lists, requires prior specific 1
permission and/or a fee. Note for example that, in large urban areas, is standard
IWCMC’09, June 21–24, 2009, Leipzig, Germany. practise to install antennas of mobile operators on top of
Copyright ° c 2009 ACM 978-1-60558-569-7/09/06 ...$5.00. residential buildings.
 Application
VoIP Announc Missing Messaging

Middleware
Storage Pub/Sub Location
B
Context

Transport
TCP UDP RTP ...

Route Selection

Routing
Ad Hoc IP Mesh
C

Physical/Link
802.11 802.16 DSL UMTS Prop

Figure 1: Application examples of the Backup Net-

work Figure 2: Components of the Backup Network

networks, our network is self-manageable and therefore does using an access point in the neighbourhood. In our idealised
not represent an additional burden to the participants. network, a myriad of other combinations of the different
The contributions of a backup network would be two-fold: medias involved is possible.
i) emergency personnel would benefit of an extended cover- Although most of the hardware required to achieve such
age, improving the quality and amount of data provided and a generic and transparent network is already in place, the
received from the command centre; and ii) civilians would deployment of the network poses a number of challenges
have an additional mechanism to make emergency calls. that have still to be resolved and which are the focus of the
following section.
2. OVERVIEW
The Backup Network we propose is formed by heteroge- 3. CHALLENGES
neous devices, the majority of them made available by civil- The Backup Network leverages on a number of well-known
ians. Examples are smart phones, PDA’s, laptops and do- and established networking technologies. However, most of
mestic or institutional WiFi access points (e.g. located at these technologies have so far been developed and used sep-
public schools or universities). Rescue teams contribute with arately. Therefore, we anticipate that the main challenges
their own personal communication equipment and with mo- will arise from their integration. To facilitate this integra-
bile base stations, integrated in emergency vehicles. Private tion, we propose to rely on the overall architecture of the
operators contribute by making their infrastructure avail- network depicted in Figure 2.
able, possibly relaxing access constraints. The architecture is divided in five layers, with most of
In our architecture, this myriad of devices and technolo- them mirroring the standard TCP/IP protocol composition
gies is integrated in a coherent data network by specialised stack. In brief, each layer will be composed of multiple pro-
software, made publicly available in advance, for example at tocols and software to provide their integration. In most
the web site of the national civil protection authority. The of the devices, only a subset of these protocols, those rel-
software remains deactivated by default and is manually ac- evant for the network technologies supported, will be im-
tivated by the users in response to the catastrophe. plemented. The most significant challenges are expected to
The network leverages on well-known protocols for ad-hoc, appear at the routing and middleware layers as these en-
mesh, delay tolerant and wired networks. Devices coopera- capsulate the key functionality of the various networking
tively learn the routes available between the command cen- technologies to be integrated.
tre and the participants. These routes may traverse different In the following subsections, we discuss the challenges
networking technologies. Therefore, devices simultaneously posed to each layer and outline the principal components
accessing multiple networks are required to act as gateways. that are envisioned to resolve them.
Routes are selected according to the available quality of
service. In the general case, the goal of the nodes in the 3.1 Physical and Link Layer
Backup Network is to deliver packets to the more “stable” To be useful, the Backup Network must support a num-
parts of the network, that is, to the locations where either ber of physical and link layer protocols. Wireless proto-
the civil or rescue networks are operational. cols include IEEE 802.11 (WiFi), IEEE 802.16 (WiMAX),
Figure 1 illustrates 3 examples of operation of the Backup GSM/GPRS, UMTS and possibly some proprietary protocol
Network. Case A illustrates the establishment of a voice call used on the rescue network. The network is also expected
using a smart phone. Since the civil network infrastructure to access wired protocols like DSL and Ethernet. To facili-
of the region was affected, communication is mediated by tate, the location of the resources available to the network,
another smart phone and a mobile base station that connects participants must agree on a unique, link layer independent,
to the Rescue Network. Mediation also takes place for the network ID.
establishment of the call in case B. However, in this case the Because of the extraordinary circumstances we are consid-
connection uses the civil network. Finally, case C depicts an ering, we expect that limitations of associations between de-
example where, due to the local failure of the rescue network, vices and infra-structured networks to be alleviated. In the
a passing by rescue vehicle communicates its current location case of WiFi networks, this includes the dismiss of any secu-
rity mechanisms. In mobile telephones, telecommunication Packet Prioritisation. The bandwidth differences in the
companies are expected to temporary drop the restrictions networks traversed by packets in transit is likely to produce
of use of clients from other operators or from those with their traffic accumulation at the nodes. The Backup Network
service suspended. Section 3.5 discusses some measures to must provide the means to prioritise traffic. The contribu-
prevent abuses of the Backup Network. tion of priorities is two-fold: i) they allow the network to
It is common to use any of these wireless protocols for privilege traffic produced or directed to rescue teams; and
data delivery. The challenge resides in their integration. ii) they assure the rescue network is used for the purpose it
Link layer bridging is expected to be performed by any de- was initially designed for so that other traffic, produced by
vice with two or more network interfaces, namely, access civilians, does not interfere with rescue operations.
points, base stations and smart phones. The later are ex- Packet Tagging. Route selection can also be influenced
pected to present a significant challenge because the multi- by the type of data carried in packets. The Backup Network
ple interfaces provided by smart phones are aimed to serve should therefore provide a packet-level tagging mechanism
exclusively as alternative communication points for appli- that includes potential Quality of Service attributes. For
cations. Most likely, the design of smart phones will not example, a Voice Over IP (VoIP) conversation should not
consider problems like concurrent packet reception. not be forwarded over a network that is subject to large la-
tencies like a Delay Tolerant Network. On the other hand,
3.2 Network Layer batch application messages, like updates to the location of
the emergency vehicles can suffer some delay and jitter with-
The role of the network layer is to drive data packets to
out severely affecting the service they support.
the destination. In the Backup Network, a number of unique
Network Status. Collecting all the information that will
challenges are posed to this layer.
assist route selection is challenging, given that one must con-
Addressing. The unstructured model of the Backup Net-
sider functional and non-functional requirements and that
work conflicts with the structured IP addressing scheme.
most of this information can not be inferred locally by each
However, the infra-structured components of the Backup
device. Instead, devices must agree on a protocol to con-
Network are likely to use the ubiquitous Internet protocols.
struct an up-to-date distributed view of the network and
Therefore, the Backup Network must provide to all partici-
the devices it contains.
pants some schema permitting to acquire IP addresses. Re-
An accurate view of the Backup Network status allows, for
cent research results propose to delegate on the gateways
example, to identify working connection points with Civilian
between the infra-structured and the ad-hoc components of
and Rescue Networks. Such a separation can be used to limit
the network the role of DHCP servers [17, 14]. The solution
the use of the Rescue Network by civilians exclusively as a
is not straightforward and two problems must be resolved: i)
last resort; to avoid routes traversing congested regions or
confirming the uniqueness of IP addresses consumes a non-
to avoid mobile devices with low battery.
negligible amount of resources and; ii) depending on the
available connectivity and node movement, mobile nodes
may become associated to the infrastructure on different 3.3 Middleware Layer
points, possibly changing their addresses. On federations One of the objectives of the middleware layer is to provide
of access points (e.g. on an operational rescue network) services that hide from the application the complexity of the
this problem is trivially solved using hand off protocols like underlying network. This objective gains relevance in the
HAWAII [20]. At the Internet scale, Mobile IP [18] Home highly heterogeneous environment of the Backup Network.
Agents can be used to keep the mapping of each device on its As a minimal set, the Backup Network should provide the
current (foreign) IP address. However, Mobile IP is designed following middleware services:
assuming that devices have a Home Network. To provide to Distributed Data Storage. The exchange of data plays
devices participating in the Backup Network a similar func- a key role in the kind of backup network we envisage. For
tionality, it would be necessary to define a default Home instance the command centre will need to disseminate to
Network and a corresponding Home Agent. This can raise rescue teams guidelines, maps, reports of the most critical
scalability problems due to the expectantly large number of situations, alerts for missing persons, etc. Reciprocally res-
devices that would try to associate to this network. cue teams will need to produce regular updates regarding
Routing Metrics. The majority of routing protocols, for the situation they are facing, possibly in the form of an-
both wired [16] and wireless [12, 19] networks, select the notation of shared digital documents, such as maps, or by
route with the lowest number of intermediary hops between updating the system’s collective view of the situation (e.g.
the source and the destination. In homogeneous networks, by removing a person from the list of missing persons when
the number of intermediary hops is roughly proportional to this person is found).
the packet delay. To support this kind of collaborative interactions, the mid-
In the Backup Network, the preferable route may not be dleware of the Backup Network should provide a distributed
the one with a lower number of intermediary hops. Instead, storage that is accessible to both producers and consumers
the routing protocol should privilege those routes that tend of data. This data management service should also be repli-
to be more stable and are less congested. In general, these cated, to improve its accessibility even in periods and areas
should be routes traversing infra-structured networks. How- with limited connectivity. It should also tolerate a high rate
ever, to reach the infrastructure, packets may have to be of ongoing failures, as these are bound to be present in a
driven by ad-hoc networks. That is, routing is likely to re- major catastrophe scenario.
quire vertical and horizontal hand offs. The definition of Publish/Subscribe Service. In a catastrophe situation,
novel metrics, weighting the particularities of each network many of the messages produced will have a destination that
type is therefore a critical enabler of the Backup Network is not fully specified. Instead, the destination of a message
this paper advocates. can be specified by some (possibly not formal) class of par-
ticipant in the rescue missions such as ambulance drivers or communication channel makes room for a number of addi-
members of rescue teams in proximity of some location. If a tional applications that can contribute to rapidly mitigate
communication media providing this class of addresses was the adverse effects of the catastrophe. In particular, the
available, it would serve nicely queries to the closest ambu- Backup Network opens a bidirectional communication chan-
lance available, or to advertise the locations with explosion nel between authorities and the public. In addition to emer-
danger due to gas leaks. gency calls, it can also be used to make public announce-
The properties of a publish/subscribe service (cf. [1, 3, ments that contribute to preserve public order.
6]) fit nicely in this communication model, with users sub- If conditions permit it, the network could be used for the
scribing to all the topics or attributes they find relevant. An establishment of bidirectional communication channels be-
important aspect is that in publish/subscribe, producers are tween civilians. Depending of the quality of service available,
decoupled from subscribers. This contributes to reduce the the application could establish voice communication, chats
application programing complexity and copes well with the or off-line messaging. The distributed data storage middle-
communication model of the network expected to present ware can be used for a missing persons location service.
frequent disconnections and changes in the topology. Preventing data overload at the command centre is a chal-
Context Service. Applications designed for the Backup lenge. According to newspaper reports[21, 9], in a recent
Network will be required to cope with intermittent connec- large scale simulation of an earthquake in the Lisbon urban
tivity and high variations of bandwidth, latency and jitter. area, communication was identified as a major bottleneck
In addition, applications will be developed for a high number that limited the operational capacity of the rescue teams
of brands of mobile devices and it should be impractical to and hospitals. The simulation was mostly driven by the civil
configure each brand and model individually. The context protection authorities and should have not considered the
service facilitates applications self-adaptation by gathering, immense number of calls to emergency numbers that would
summarising and interpreting context information. have occurred in a real situation. Large scale message tag-
Location Service. Accurate location is fundamental for ging, detection of duplicates and appropriate filtering and
an efficient management of the resources and for the rapid routing are fundamental criteria for a successful manage-
localisation of the events. However, it should not be ex- ment of a crisis situation. A first level of filtering could be
pected that all participating devices are equipped with GPS applied on applications deployed upstream of the command
receivers. The role of this service is to allow any device to centre, for example in rescue vehicles, thus reducing message
retrieve its location with a reasonable accuracy. This service flow up to the command centre.
will not represent any burden to the infrastructure. Devices In complement, a data concentration application at the
not equipped with GPS receivers can estimate their location command centre would refine event filtering and tagging,
from the information provided by nodes in the neighbour- ensuring a scalable and manageable flow of information.
hood.
Implementing the above services in the context of a major 3.5 Security and Privacy Considerations
catastrophe is particularly difficult. As mentioned above, Deployment of this communication media would be mostly
the middleware will have to face extreme levels of hetero- performed in advance of the catastrophe using trusted web
geneity, in terms of devices, bandwidth, connectivity, ro- sites allowing users to download and install the required ap-
bustness. Our general philosophy consists in relying on plications and middleware. Due to privacy and security is-
decentralised schemes to reduce the load at the command sues, users are invited to activate the application only upon
centre and to alleviate the traffic on the infrastructure. In the occurrence of the catastrophe. Therefore, the impact of
particular, epidemic protocols (e.g. [4, 10, 8, 23]), and mid- the application would be restricted to the memory consumed
dleware based on them [15] seem particularly promising to on the devices. Because the application is not operational,
provide both a high level of resilience, and the kind of self- no privacy concerns are raised. In operation, preventing
organisation required in an adverse and unpredictable envi- devices from leaking personal information requires the con-
ronment. Key challenges will be to trade-off between con- sumption of additional resources of the devices, which would
flicting requirements, and understand how to use epidemic be undesirable. However, we expect users to relax their pri-
protocols as a gluing technology between heterogeneous and vacy concerns in the aftermath of a catastrophe.
possibly damaged underlying infrastructure (WiFi, GSM, Due to its open nature, the Backup Network is vulnerable
dedicated civil and military networks, etc.). to intentional and unintentional attacks. An interesting as-
In terms of architecture, the Backup Network can be per- pect of this model is that problems can be contained in each
ceived to be a system of systems (or a network of net- ad-hoc component of the network by requesting gateways
works) [22]. In this context existing middleware architec- to filter the traffic forwarded to the infrastructure compo-
tures such as Ambient Networks [25], Open Overlays [7] or nent. Gateways can for example, drop traffic not addressed
SpoVNet [26], seem particularly promising to support the in- to the command centre (thus prohibiting calls between civil-
tegration of the heterogeneous systems. The challenge again ians) or impose a limit on the number of packets forwarded
will consist of adapting these approaches to the particular to the command centre, contributing to the mitigation of a
requirements of a Backup Network, and provide a close in- potential DDoS attack.
tegration at the middleware level between QoS, robustness,
and scalability under adverse and unpredictable conditions. 4. CONCLUSIONS AND FUTURE WORK
Reliable and pervasive communications play a fundamen-
3.4 Applications tal role in the coordination of the rescue efforts after a ma-
The primary role of the Backup Network is to provide jor catastrophe. Unfortunately, in large scale events, one
an alternative communication channel between rescue teams cannot assume the complete resilience of the infrastructure.
and the command centre. However, the availability of this This position paper outlines the challenges that must be ad-
dressed for the establishment
View publication stats
of a backup network, aiming Networks, pages 139–172. Addison-Wesley, 2001.
to mitigate localised failures in the infrastructure. The net- [13] K. Kanchanasut, A. Tunpan, M. Awal, D. Das,
work is created by the off-the-shelf networking equipment T. Wongsaardsakul, and Y. Tsuchimoto.
that is now ubiquitous in any European city. Examples are DUMBONET: a multimedia communication system
smart phones, laptops and access points. These devices co- for collaborative emergency response operations in
operate to establish the first hop of a network providing disaster-affected areas. Int’l Journal of Emergency
access to the infrastructure that survived the catastrophe. Management, 4(4):670–681, 2007.
The primary goal of this network is to mitigate failures in [14] S.-C. Kim and J.-M. Chung. Message complexity
communication of rescue teams. However, a number of ad- analysis of mobile ad hoc network address
ditional applications, possibly used by civilians, can also be autoconfiguration protocols. Trans. on Mobile
considered. Computing, 7(3):358–371, 2008.
In the near future, the authors plan to investigate these [15] S. Lin, F. Taı̈ani, and G. Blair. Facilitating gossip
challenges in detail and perform small scale localised exper- programming with the gossipkit framework. In 8th
iments of integration of the myriad of technologies involved. IFIP Conf. on Distributed Applications and
Interoperable Systems, 2008.
5. REFERENCES [16] G. Malkin. RIP version 2. RFC 2453, IETF, 1998.
[1] R. Baldoni, R. Beraldi, G. Cugola, M. Migliavacca, [17] A. Misra, S. Das, A. Mcauley, and S. Das.
and L. Querzoni. Structure-less content-based routing Autoconfiguration, registration and mobility
in mobile ad hoc networks. In Int’l Conf. on Pervasive management for pervasive computing. Personal
Services (ICPS’05), pages 37–46, 2005. Communications Systems Magazine, 8:24–31, Aug.
[2] P. Costa, G. Coulson, C. Mascolo, G. Picco, and 2001.
S. Zachariadis. The RUNES middleware: a [18] C. Perkins. IP mobility support for IPv4. RFC 3344,
reconfigurable component-based approach to IETF, 2002.
networked embedded systems. In 16th Int’l Symp. on [19] C. Perkins and E. Royer. Ad-hoc on-demand distance
Personal, Indoor and Mobile Radio Comm. (PIMRC vector routing. In 2nd IEEE Workshop on Mobile
2005), volume 2, pages 806–810, 2005. Computing Systems and Applications, pages 90–100,
[3] G. Cugola, E. D. Nitto, and A. Fuggetta. The JEDI 1999.
event-based infrastructure and its application to the [20] R. Ramjee, T. L. Porta, S. Thuel, K. Varadhan, and
development of the OPSS WFMS. Trans. on Software S. Wang. HAWAII: A domain-based approach for
Engineering, 27(9):827–850, 2001. supporting mobility in wide-area wireless networks. In
[4] A. Demers, D. Greene, C. Hauser, and et al. Epidemic 7th Int’l Conf. on Network Protocols, pages 283–292,
algorithms for replicated database maintenance. In 1999.
Symp. on Principles of Distributed Computing, 1987. [21] RTP. Sismo/simulacro: Dificuldades nas comunicações
[5] O. Drugan, T. Plagemann, and E. Munthe-Kaas. encontradas no inı́cio do exercı́cio já foram
Resource aware middleware services over manets. In ultrapassadas. http://ww1.rtp.pt/noticias/index.php?-
25th Int’l Conf. on Computer Comm. (INFOCOM article=374301&visual=26&tema=1, Nov. 21 2008. In
2006), pages 1–2, 2006. portuguese.
[6] P. Eugster, P. Felber, R. Guerraoui, and A.-M. [22] A. Sage and C. Cuppan. On the systems engineering
Kermarrec. The many faces of publish/subscribe. and management of systems of systems and
Computing Surveys (CSUR), 35(2):114–131, 2003. federations of systems. Information, Knowledge,
[7] P. Grace, D. Hughes, B. Porter, G. Blair, G. Coulson, Systems Management, 2(4):325–345, 2001.
and F. Taiani. Experiences with open overlays: a [23] Y. Sasson, D. Cavin, and A. Schiper. Probabilistic
middleware approach to network heterogeneity. In 3rd broadcast for flooding in wireless mobile adhoc
SIGOPS/EuroSys European Conference on Computer networks. Wireless Comm. and Networking, 2003.
Systems (Eurosys ’08), pages 123–136, 2008. [24] Y. Shibata and K. Takahata. Performance evaluation
[8] Z. Haas, J. Halpern, and L. Li. Gossip-based ad-hoc of large scale disaster information network. In Int’l
routing. In Trans. on Networking, 2006. Conf. on Parallel Processing Works. (ICPPW’07),
[9] IOL Diário. Simulacro de sismo revelou “fragilidades” page 7, 2007.
nos meios. http://diario.iol.pt/sociedade/sismo- [25] M. Stiemerling et al. System Design of SATO and ASI.
simulacro-proteccao-civil-lisboa/1016368-4071.html, Deliverable D12-F.1, Ambient Networks Project, 2006.
Nov. 23 2008. In portuguese. [26] W. Waldhorst, C. Blankenhorn, D. Haage, R. Holz,
[10] M. Jelasity and O. Babaoglu. T-man: Gossip-based G. Koch, B. Koldehofe, F. Lampi, C. Mayer, and
overlay topology management. In Engineering S. Mies. Spontaneous virtual networks: On the road
Self-Organising Systems, 2005. towards the internet’s next generation. it- Information
[11] X. Jiang, N. Chen, J. Hong, K. Wang, L. Takayama, Technology, Dec. 2008.
and J. A. L. Siren: Context-aware computing for
firefighting. In 2nd Int’l Conf. on Pervasive
Computing (Pervasive 2004), volume 3001 of Lecture
Notes in Computer Science, pages 87–105, 2004.
[12] D. Johnson, D. Maltz, and J. Broch. Ad Hoc
Networking, chapter DSR: The Dynamic Source
Routing Protocol for Multi-Hop Wireless Ad Hoc