Top Banner
Dynamic IP configuration of terminals in broadcasting networks G. Gardikis a, * , S. Orfanos a , G. Kormentzas a , E. Pallis c , A. Kourtis b a University of the Aegean, Department of Information and Communication Systems Engineering, 83 200 Karlovassi, Samos, Greece b National Center for Scientific Research ‘Demokritos’, Institute of Informatics and Telecommunications, Patriarchou Gregoriou strasse, Ag. Paraskevi, Athens 165 10, Greece c Department of Applied Informatics and Multimedia, Technological Educational Institute of Crete, Estauromenos, 715 00 Heraklion, Crete, Greece Available online 8 September 2007 Abstract The concept of synergy between broadcasting and telecommunication networks has been strengthened by the emer- gence of multi-modal terminals, which are used in a broadcast environment (mainly in DTV-Digital Television networks) to provide IP-based multimedia services. The migration of IPv4/IPv6 applications, either interactive or not, in a broadcast- ing network, requires that certain parameters, such as Host, Gateway and DNS IP addresses are configured in the termi- nals, either statically or dynamically. This paper discusses issues of dynamic configuration of IP parameters for DTV terminals, based on an overview of relevant mechanisms usually used in access networks. It proposes an IP-based auto- configuration protocol tailored to the needs of an IP/DTV access platform, describes its implementation and evaluates its behaviour in a laboratory-based DVB-T network. Ó 2007 Elsevier B.V. All rights reserved. Keywords: IP-over-DTV; Interactive broadcasting; Dynamic configuration 1. Introduction The worldwide tendency for global convergence of networks and services has resulted in the trans- formation of digital broadcast networks, from mere carriers of unidirectional media streams to broad- band access networks for integrated, all-IP services. Contemporary digital television (DTV) networks, as successors to analog TV systems, are being used as access networks for IP-based services, either interac- tive or not. The common digital baseband format, namely the MPEG-2 Transport Stream, previously used only for the transport of encoded audio/visual streams, has extended its functionality to support all types of digital information, including IP data (‘‘Datacasting’’). The adoption of adaptation proto- cols such as multi-protocol encapsulation [1] and unidirectional lightweight encapsulation [2] for the insertion of IP data into the MPEG-2 TS, has greatly contributed to the migration of native IPv4/IPv6 applications to the world of digital broadcasting, both satellite and terrestrial. The use of broadcasting terminals for accessing IP services requires that they incorporate an IPv4/IPv6 protocol stack. For the proper operation of the lat- ter, it is essential that certain parameters have to be 1389-1286/$ - see front matter Ó 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.comnet.2007.09.006 * Corresponding author. Tel.: +30 210 6503108; fax: +30 210 6532175. E-mail address: [email protected] (G. Gardikis). Available online at www.sciencedirect.com Computer Networks 52 (2008) 292–302 www.elsevier.com/locate/comnet
11

Dynamic IP configuration of terminals in broadcasting networks

Apr 20, 2023

Download

Documents

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Dynamic IP configuration of terminals in broadcasting networks

Available online at www.sciencedirect.com

Computer Networks 52 (2008) 292–302

www.elsevier.com/locate/comnet

Dynamic IP configuration of terminals in broadcasting networks

G. Gardikis a,*, S. Orfanos a, G. Kormentzas a, E. Pallis c, A. Kourtis b

a University of the Aegean, Department of Information and Communication Systems Engineering, 83 200 Karlovassi, Samos, Greeceb National Center for Scientific Research ‘Demokritos’, Institute of Informatics and Telecommunications,

Patriarchou Gregoriou strasse, Ag. Paraskevi, Athens 165 10, Greecec Department of Applied Informatics and Multimedia, Technological Educational Institute of Crete, Estauromenos,

715 00 Heraklion, Crete, Greece

Available online 8 September 2007

Abstract

The concept of synergy between broadcasting and telecommunication networks has been strengthened by the emer-gence of multi-modal terminals, which are used in a broadcast environment (mainly in DTV-Digital Television networks)to provide IP-based multimedia services. The migration of IPv4/IPv6 applications, either interactive or not, in a broadcast-ing network, requires that certain parameters, such as Host, Gateway and DNS IP addresses are configured in the termi-nals, either statically or dynamically. This paper discusses issues of dynamic configuration of IP parameters for DTVterminals, based on an overview of relevant mechanisms usually used in access networks. It proposes an IP-based auto-configuration protocol tailored to the needs of an IP/DTV access platform, describes its implementation and evaluatesits behaviour in a laboratory-based DVB-T network.� 2007 Elsevier B.V. All rights reserved.

Keywords: IP-over-DTV; Interactive broadcasting; Dynamic configuration

1. Introduction

The worldwide tendency for global convergenceof networks and services has resulted in the trans-formation of digital broadcast networks, from merecarriers of unidirectional media streams to broad-band access networks for integrated, all-IP services.Contemporary digital television (DTV) networks, assuccessors to analog TV systems, are being used asaccess networks for IP-based services, either interac-tive or not. The common digital baseband format,

1389-1286/$ - see front matter � 2007 Elsevier B.V. All rights reserved

doi:10.1016/j.comnet.2007.09.006

* Corresponding author. Tel.: +30 210 6503108; fax: +30 2106532175.

E-mail address: [email protected] (G. Gardikis).

namely the MPEG-2 Transport Stream, previouslyused only for the transport of encoded audio/visualstreams, has extended its functionality to support alltypes of digital information, including IP data(‘‘Datacasting’’). The adoption of adaptation proto-cols such as multi-protocol encapsulation [1] andunidirectional lightweight encapsulation [2] for theinsertion of IP data into the MPEG-2 TS, hasgreatly contributed to the migration of nativeIPv4/IPv6 applications to the world of digitalbroadcasting, both satellite and terrestrial.

The use of broadcasting terminals for accessing IPservices requires that they incorporate an IPv4/IPv6protocol stack. For the proper operation of the lat-ter, it is essential that certain parameters have to be

.

Page 2: Dynamic IP configuration of terminals in broadcasting networks

Fig. 1. IP/DTV access scenarios depending on the type of returnchannel: (a) unidirectional, (b) bi-directional with native DTVinteractivity and (c) hybrid bi-directional with an externalinteraction network.

G. Gardikis et al. / Computer Networks 52 (2008) 292–302 293

configured, such as IPv4/IPv6 host, gateway andDNS addresses. These parameters have been untilnow set in a static manner, a rational approach ifone considers that IP/DTV service provision ismainly restricted to the satellite access domain(e.g., via DVB-S/DVB-RCS platforms), where fixedterminals almost never require re-configuration.Nevertheless, with the emergence of terrestrial digitaltelevision infrastructures, where mobile use fromportable terminals aiming at ubiquitous access isanticipated [3], a dynamic/automatic mechanism IPconfiguration is essential. In this context, this paperdiscusses the issue of dynamic IP set-up of interactiveDTV/IP terminals. It shows why widely used mech-anisms like DHCP are inadequate due to the specialfeatures of digital television networks, and proposesand implements an IP-based protocol approach.

The article proceeds as follows: Section 2 pre-sents a general topology of IP/DTV networks anddivides them into three rough categories, outliningthe configuration needs of each case. Section 3 dis-cusses the requirements that an IP auto-configura-tion mechanism should fulfil in a DTV networkand investigates whether existing solutions, such asDHCP, could be used. Section 4 proposes a novelapproach for an IP-based dynamic configurationprotocol, describing its functionality and suggestinga state diagram. Section 5 presents the implementa-tion of the protocol into platform-independent Javamodules and evaluates its behaviour in a labora-tory-based interactive DVB-T set-up. Finally, Sec-tion 6 concludes the paper.

2. IP-over-DTV access scenarios

For simplicity reasons, in this article, the term‘‘Digital Television Network’’ is assumed to referto a single DTV downlink conveying a singleMPEG-2 TS to a certain coverage area (e.g., a satel-lite footprint). However, the modelling and themechanisms described in this paper can be easilyexpanded to include multi-bouquet, multi-transmit-ter infrastructures.

At the broadcaster’s site, an IP-to-DTV encap-sulator is used to insert IPv4/IPv6 traffic in theMPEG-2 Transport Stream. On the other side,the terminal, following demodulation, decodingand demultiplexing, decapsulates the datacaststreams and feeds them to its IP stack for process-ing. Since the digital broadcast channel is unidirec-tional, the support for fully interactive servicesrequires the presence of an interaction channel

[4]. The existence and nature of the latter is crucialto the IP configuration process and, in this context,the DTV network topologies can be categorised

Page 3: Dynamic IP configuration of terminals in broadcasting networks

294 G. Gardikis et al. / Computer Networks 52 (2008) 292–302

into three discrete generalised architectures (seeFig. 1).

2.1. Unidirectional topology

This case is the most usual nowadays andincludes all broadcast-only platforms, satellite orterrestrial, where no return link (from the terminalback to the broadcaster) exists. The terminalsoperate in a receive-only mode and receive DTVprograms multiplexed with datacast IP data. Dataservices are not fully interactive, they can only pro-vide pseudo-interactivity by means of off-line localaccess. In this case, the terminals have no way tointeract with a network entity at the broadcaster’sside to request configuration information. The lat-ter can only be ‘‘pushed’’ to all receivers. DNS andgateway addresses are of no use, and a common IPhost address can be assigned to all terminals forfunctionality reasons only. Since the DTV receiverscannot send data, this common address does notraise any conflict issues.

2.2. Bi-directional topology with native DTV

interactivity

In this case, the terminals support full interactiv-ity by utilising a unidirectional interaction (return)channel that has been specially developed to inte-grate with DTV networks, such as DVB-RCS(Return Channel via Satellite) or DVB-RCT(Return Channel-Terrestrial). Unlike the previousscenario, it is essential in this case that the terminals’IP parameters are fully configured and their hostaddresses are unique. Until now, native DTV inter-active solutions [5] are commonly based on the sta-tic configuration approach, i.e., the terminal comespre-configured by the installer/service provider.This solution is satisfactory for certain cases, e.g.,where a satellite terminal is fixed (no handovers)and bound to a single service provider. However,for moving terrestrial or satellite transceivers per-forming handovers or in cases that multiple serviceproviders are to be supported, a dynamic solutionis essential. It is possible, for instance, that channelzapping in a terrestrial receiver could correspond toswitching among different broadcasters/service pro-viders. In order that the receiver’s settings are set tothe IP address space of each provider, an IP config-uration procedure could need to be activated aftereach zap.

2.3. Bi-directional topology with an external

interaction network

This scenario corresponds to a hybrid interactivetopology [6] where the DTV broadcast channel isused for forward datacasting, and a foreign IP net-work is utilized to support interaction. The lattercan be any wired or wireless IP-enabled infrastruc-ture, such as GPRS/3G, WLAN, PSTN or xDSL.The terminal is equipped with a dual-interface frontend: one for the broadcast network (i.e., a DTV recei-ver) and one for the interaction channel (e.g., a 3Gmodem or a WLAN interface). In this scenario, theinterface bound to the interaction network receivesthe IP configuration from its own infrastructure viaa standardized mechanism anyway (e.g., via DHCPfor a WLAN interface). However, if the broadcasteris to provide interactive services, it is essential that theDTV receiver module must also be configured.Otherwise, the broadcaster will have to deal with alarge pool of heterogeneous IP addresses, differentaccording to the interaction network provider usedby each client, and handling of interaction data willbe rather complicated. So, an IP/DTV dynamic con-figuration mechanism is also required in this case.

3. Requirements for an IP/DTV dynamic

configuration protocol

DHCP (Dynamic Host Configuration Protocol)[7], and DHCPv6 [8] are the most common IETF-standardized protocols for dynamic configurationof host IPv4 and IPv6 parameters, respectively.These protocols are used in a huge spectrum ofIP-enabled networks.

DHCP works on a client/server basis. Upon theconnection of a new host to a certain sub-network,the unconfigured host (client) broadcasts a ‘‘serverdiscovery’’ (DHCPDISCOVER) message using anetwork- and link-level broadcast address. TheDHCP Server, normally unique within a subnet,responds with a unicast DHCPOFFER message,containing the IP configuration parameters and alease period. The client responds with a DHCPRE-QUEST back to the server, requesting the offeredparameters, and the four-way handshake processends with the transmission of DHCPACK fromthe Server to the Client, confirming that the clienthas been associated with the new parameters. Theoperation of DHCPv6 is quite similar, with addedsupport for IPv6 and exploitation of its newfeatures.

Page 4: Dynamic IP configuration of terminals in broadcasting networks

G. Gardikis et al. / Computer Networks 52 (2008) 292–302 295

Unfortunately, IP/DTV networks, due to theirparticular nature, have certain requirements thatcannot be satisfied by the DHCP/DHCPv6 protocol‘‘as-is’’. The most important ones are:

• Support of all three generic topologies described in(II) via a unified approach.

• IP-unicast-only requests. Due to the heterogene-ity of a composite interactive DTV network,there is no common link-layer over whichrequests and responses could be broadcast. Thecommunication from the terminal to the broad-caster should be entirely based on unicast IPand should be independent from any underlyinglink technologies.

• Scalability. Unlike a LAN or even a cellular net-work, a DTV platform should be capable ofaccommodating even hundreds of thousands ofusers. The proposed mechanism must be scalableenough to support a large pool of end users.

• Simultaneous IPv4 and IPv6 support. The param-eters, which are dynamically configured, shouldcorrespond to both versions of the Internet Pro-tocol. This means, e.g., that the client terminalis assigned at the same time both a v4 and a v6global unicast host address via the same proto-col. Stateless IPv6 auto-configuration also hasto be supported.

• Authenticated access. Since, unlike a wired LAN,a broadcast network is ‘‘open’’ to anyone, securityfrom unauthorised requests is a major issue. Theusage of a standardized authentication methodsuch as the one described in [9] can be considered.

• Utilisation of native DTV signalling so thatbroadcast configuration data can be located inthe broadcast MPEG-2 TS.

• Relatively low complexity, so that (i) the protocolcan be implemented even in handheld terminalsand (ii) the computational cost at the broadcasterdue to the large number of request processes canbe minimised. There are many options of DHCP,such as Relaying and Solicitation, which are ofno use in either of the aforementioned scenarios.

• Easy and modular implementation on existingsatellite/terrestrial/handheld digital television ter-minals without modifications in their hardwareor their lower layers (demultiplexing/decapsula-tion modules).

It is obvious that the DHCP protocols, whichexist and are used in networks with a commonlink-layer (e.g., in an Ethernet or a WLAN) cannot

be used ‘‘as-is’’ in our case, where the communica-tion between the terminal and the configuration ser-ver cannot be established in a link-layer basis, sincethey do not belong to a common link. This holdsespecially in the case where the interaction channelis implemented by an external IP network. A newmechanism needs to be introduced, based on thephilosophy of existing dynamic configuration proto-cols, but implemented under a different perspective.

4. A proposed IP-based dynamic configuration

mechanism

The paper proposes an IP-based mechanism fordynamic configuration of the digital broadcastingterminals, which fulfills all the requirements statedin the previous section. For reference reasons, thismechanism/protocol will be referred to as IDDCP(IP-over-DTV Dynamic Configuration Protocol).This section presents the principles of operation,the message types, the algorithm and the state dia-gram of the IDDCP protocol. The philosophy fol-lowed and the message types are mainly derivedfrom the operation of DHCPv4/DHCPv6. Never-theless, many adaptations have been introduced tomatch the requirements described in the previoussection.

The approach is based on a client–server model,where all IDDCP messages are conveyed overUDP/IP, either over the broadcast channel or theinteraction network. The IP-based model allowsfor easy, IP-level implementation on any existingterrestrial or satellite DTV terminal, without anymodification in its RF/demultiplexing/decapsula-tion stack. The ‘‘Client’’ is considered to be theDTV terminal whose parameters are to be set, andthe ‘‘Server’’ is a separate entity located in thebroadcaster’s premises. The role of the IDDCP Ser-ver will be the management and allocation of IPparameters to terminals. Since the number of Cli-ents can be in the order of tens or thousands, itwould be convenient to use an open database(e.g., SQL-based) to store the assigned parametersrather than an internal table, as most DHCP Serversdo. Regarding the UDP ports used, the numbersused by DHCPv6 could be adopted: 546 for the Cli-ent and 547 for the Server.

The message types, which are required are:

• ADVERTISE: It is broadcast at regular intervalsby the Server to all Clients over the DTV channeland contains its own IPv4/IPv6 global address

Page 5: Dynamic IP configuration of terminals in broadcasting networks

Fig. 2. IDDCP state diagram (Client).

296 G. Gardikis et al. / Computer Networks 52 (2008) 292–302

(or multiple addresses if multiple Servers exist), asubnet prefix/mask for all Clients and a commontemporary IP address, which is used by all Cli-ents during the configuration procedure.

• REQUEST: It is sent by the Client to the Serverover the interaction channel to request configura-tion parameters. It contains the MAC address ofthe DTV interface and a UID (Unique Identi-fier), similar to DHCPv6’s DUID, which is usedby the Server to uniquely identify Clients.

• REPLY: It is sent by the Server via the broadcastchannel as a response to a REQUEST command.It contains a set of configuration parameters:IPv4 and IPv6 addresses of the network Gate-way, the DNS Server and of course of the Clientterminal itself. It also contains a ‘‘lease’’ value,i.e., a time duration after which the parametersassigned are invalid and they need to be renewedvia a RENEW message (see below).

• CONFIRM: It is sent by the Client to the Serveras a confirmation that the REPLY message hasbeen promptly received and accepted.

• RENEW: It is sent by the Client upon the expira-tion of the lease timer, as a request to the Serverin order to renew the assigned parameters.

• RECONFIGURE: It is sent by the Server to theClient if any of the parameters assigned need tobe changed. Upon reception of RECONFIG-URE, the Client has to re-initiate the dynamicconfiguration process.

For simplicity reasons, it is assumed that, for afirst approach, neither the Client’s request nor theServer’s offer can be declined. If this is so (an unli-kely case), then the appropriate message types willhave to be added.

The state diagram of the proposed mechanism,from the Client side, is depicted in Fig. 2. The Ser-ver-side diagram can be directly derived, and is thusomitted.

IDDCP is based on a four-way handshake. Itsoperation is described as follows:

At first, a Client, which needs to be configured(e.g., after power-up, after a handover, or when IPservices are activated) listens to the broadcaststream to locate a Server ADVERTISE message.In general, it would be convenient that a dedicatedDTV AC (AutoConfiguration) Program Number(corresponding to a certain AC PID) is devoted tothe transport of IDDCP messages. This ProgramNumber can be declared via native DTV signalling.For example, in the case of DVB networks, the stan-

dardised INT (IP/MAC Notification Table) [1] canbe used, where the IDDCP Program can be declaredas having a target_IP of 255.255.255.255 and a tar-get_MAC of 0xff:ff:ff:ff:ff:ff (broadcast) so that it isprocessed by all terminals. ADVERTISE will becontained in a UDP packet, having the broadcastIP address in its ‘‘destination’’ field.

By parsing ADVERTISE, the Client specifies theServer’s IP, the subnet prefix/mask and assigns tothe DTV interface a common_IP address, which istemporary and is the same for all configuring termi-nals. In the unidirectional case with ‘‘receive-only’’terminals, this common_IP is sufficient as describedin Section 2 and the process ends here. In the case ofIPv6, stateless auto-configuration can take place,using the advertised prefix.

In the case that a return channel exists (cases (b)and (c) of Section 2), this temporary address is usedby the Client to form the REQUEST message,which will also contain its own MAC address anda globally unique UID. The latter can be locallygenerated as described in [8]. These two values, theMAC address and the UID are used to uniquelyidentify the terminal throughout the entire IDDCPprocess. Upon sending REQUEST, the Client initi-alises and starts a timer. If the timer expires and no

Page 6: Dynamic IP configuration of terminals in broadcasting networks

G. Gardikis et al. / Computer Networks 52 (2008) 292–302 297

reply is received, it is assumed that the message waslost and it is re-sent.

The Client listens to the AC Program and waitsfor a REPLY message from the Server. REPLY issent also with a broadcast ‘‘destination IP’’address, since the Client parameters are not yetset. Among the messages carried in the program,the Client can locate its own REPLY by matchingthe UID found in the body of the message with itsown. It is recommended that for security reasons,REPLY messages do not contain the MACaddresses of the Client. If this were done, an eaves-dropper could easily record the addresses of allconfigured Clients, and could possibly use themto generate unauthorised REQUESTs. Whenreceived, the REPLY will contain all necessaryinformation, as aforementioned, and the Clientuses them to configure itself.

When configuration is complete, the Client willrespond with a CONFIRM message, now using itsnew IP address. If a duplicate REPLY is received,it is assumed that CONFIRM was lost and it is re-sent. At the same time, a lease timer is initialised,as declared in REPLY. The configuration is consid-ered valid until the lease expires. When this happens,the Client will have to send a RENEW message tothe Server and receive another REPLY with thesame or updated configuration information.

Finally, the Server will have the option of send-ing a (unicast or broadcast) RECONFIGURE mes-sage to the Client(s), to force them to re-initiate thedynamic configuration process before the leaseexpires.

Fig. 3. Laboratory testbed for the validation

After the completion of the process, the Client isin the BOUND state, and the IP parameters of itsDTV interface are properly set. The terminal hasnow been enabled for interactive IP-based DTVservices.

The aforementioned procedure is an initial pro-posed approach to the required dynamic configura-tion protocol and can be further enhanced tosupport additional functionality, such as authenti-cation and security. However, it has to be kept atrelatively low complexity so that it can be easilyimplemented even in handheld terminals and so thatcomputational demands on the Server are kept atreasonable levels.

5. Implementation and evaluation

In order to validate the proposed configurationmechanism, a laboratory-based, yet fully functionaltestbed was implemented, as depicted in Fig. 3.

The architecture is based on DVB-T as broadcastplatform, and on a WiFi network providing theuplink (return channel). The DVB-T chain (IP-to-DVB Encapsulator, Multiplexer, Modulator) isbased on commercial modules, and a low-powerRF amplifier is used for indoor transmission ofthe DVB-T signal. The return channel is realisedby a WiFi access point feeding the provider’sLAN with the user requests via a router-networkemulator. The latter runs the NISTNet [10] networkemulator to yield pre-defined amounts of delay, jit-ter and packet loss on the uplink. The wireless user

and evaluation of the IDDCP protocol.

Page 7: Dynamic IP configuration of terminals in broadcasting networks

298 G. Gardikis et al. / Computer Networks 52 (2008) 292–302

terminal is Linux-based and equipped both with aDVB-T receiver and a WiFi interface.

The IDDPC Client and Server modules wereimplemented as lightweight, stand-alone, platform-independent Java applications. The Java implemen-tation allows for portability of the Client module tovarious kinds of terminals, including portableand handheld. In the specific Linux terminal,communication with the DVB interface is achievedvia the open LinuxTV driver subsystem. TheIDDCP Server module resides on a dedicatedLinux-based workstation at the service providerside and is designed under a multi-threaded archi-tecture to accompany multiple simultaneous clientrequests.

A first test is performed with a single Client tovalidate the proper operation of the protocol. Theduration of the whole procedure (bind time) hasan average of 5 s, and at the end the DVB interfacehas been assigned new parameters for IPv4 addressand subnet, gateway address and IPv6 Global scopeaddress. This result is illustrated by viewing the out-put of ifconfig in the DVB interface of the clientbefore and after the IDDCP procedure (Fig. 4).

After the proper operation of the mechanism hasbeen validated, its behaviour can be evaluated inrelation to the condition of the access network. Spe-cifically, the impact of round-trip delay and packetloss is examined. The quantity that is measured is

Fig. 4. Impact of the IDDCP mechanism: ifconfig o

the protocol bind time, defined as the time intervalfrom the AUTOCONF to the BOUND state ofthe client (see Fig. 2).

The round-trip time is measured via ICMP Echorequests over the asymmetric connection (via theWiFi uplink to the Server and back via the DVB-T chain). In the laboratory set-up, RTT is restrictedto 70 ms, but in a large-scale deployment it couldincrease due to the congestion or the complexityof the network – mainly of the return channel.The RTT can be increased via the NISTNet module,which inserts extra delay in the packets which tra-verse the return channel.

The bind time is measured at the Client, by aninternal timer incorporated in the IDDCP Clientmodule. Apart from the propagation delay of themessages, caused by the network itself, the Clientspends some time waiting for the next ADVERTISEmessage (ADVERTISE messages are broadcastevery 1 s), for the Server’s REPLY, for the networkinterface to accept the new parameters and for theduplicate REPLY timeout to expire. The latter, asexplained, is essential so as to ensure that theCONFIRM message has not been lost (such a losscould trigger a retransmitted REPLY). All timeouttimers, which are needed to recover from lost mes-sages both in the Client and the Server, have beenset to 3 s. The first series of measurements showthe effect of network RTT – as adjusted by the

utput before and after dynamic configuration.

Page 8: Dynamic IP configuration of terminals in broadcasting networks

Fig. 5. Impact of the network round-trip time on the bind time of the IDDCP mechanism.

G. Gardikis et al. / Computer Networks 52 (2008) 292–302 299

network emulator and measured via ICMP – to thebind time of the IDDCP mechanism. The results arederived in a loss-free environment and are shown inFig. 5. Each result was calculated as an average of10 iterations. A minimum bind time of 4 s is mea-sured, although the network parameters have actu-ally been configured much earlier, before theduplicate REPLY timeout expires.

The next step is the evaluation of the protocolin a lossy environment. Since all messages areconveyed over UDP, messages can be lost due tocongestion or erroneous links. As explained, themechanism has been designed to recover from lossesvia positive acknowledgements and retransmissionprocedures.

Fig. 6. Impact of the network packet loss on t

Packet loss conditions are caused by the networkemulator, in which return-channel packets are uni-formly dropped with a given probability. Due totimeout expirations and retransmissions, the totalbind time is prolonged. Measurements were derivedfor various packet loss rates and for a fixed RTT of500 ms. Each result was calculated as an average of30 iterations. Two sets of measurements are pre-sented in Fig. 6, for timeouts of 2 and 5 s, respec-tively, both in the Client and the Server.

It is normal that a higher timeout value results inprolonged bind time for a given packet loss rate. Ina large-scale network, the timeouts of the protocolshould be set as low as possible, but sufficientlyabove the RTT of the network, so that delayed

he bind time of the IDDCP mechanism.

Page 9: Dynamic IP configuration of terminals in broadcasting networks

Fig. 7. Performance of the mechanism under a multi-user load.

300 G. Gardikis et al. / Computer Networks 52 (2008) 292–302

packets are not misinterpreted as lost. A dynamicadjustment of the timeout according to the mea-sured network delay would be a more flexible andefficient approach.

All the aforementioned tests were carried outunder a single-client scenario. In order to evaluatethe operation of the Server under multiple concur-rent requests, a multi-user scenario was employed.Up to five Client modules were used, running therequest-bind process continuously, so that at anytime the Server was processing concurrent requests.The average bind time for all Clients was againrecorded, as depicted in Fig. 7. Loss-free channelswere assumed, with a typical RTT of 500 ms.

This test validated the proper operation of themulti-threaded Server module serving multiple Cli-ents concurrently and showed that, for a restrictednumber of simultaneous requests, the overheadintroduced in the bind time is negligible.

In conclusion, the evaluation procedure of theproposed mechanism showed that the algorithm isoperable and stable, has a smooth behaviour underlarge-delay conditions, and can recover from packetloss. Multi-client operation was also validated.

6. Conclusions

This article discussed the issue of dynamic config-uration of IP parameters of terminals in broadcast-ing networks. Via an overview of the broadcastnetwork scenario, the need of such a mechanismwas justified, particularly in terrestrial networks,where handovers and switches among broadcasters

are common. Through an overview of the particu-larities of a broadcasting network and the require-ments for dynamic configuration, it was explainedwhy existing protocols such as DHCP cannot beused ‘‘as-is’’. A new client–server mechanism wasproposed, namely IDDCP (IP-over-DTV DynamicConfiguration Protocol) and was extensivelydescribed via its message types, operational descrip-tion and state diagram. This protocol is designed toprovide simultaneous IPv4/IPv6 configurationand is well tailored to suit the requirements of anIP/DTV interactive network via a simplified yetfully functional and scalable approach, which canbe easily implemented at the IP-level in all existingsatellite and terrestrial/handheld terminals. Theproposed protocol was implemented in platform-independent lightweight Java modules, and itsbehaviour was evaluated in relation to networkRTT, packet loss, and multi-Client operation. Theresearch effort referred to in this paper was carriedout in the frame of the ‘‘IP-over-DVB’’ workgroupof the IETF [11].

Acknowledgments

The concept reported in this paper was elabo-rated within the project ‘‘Study and Developmentof Interactive Broadband Services based on DVB-T/DVB-H Technologies’’ in the context of frame-work 2.2 of ‘‘Pythagoras II: Research GroupSupport of the University of the Aegean’’ jointlyfunded by the European Union and the Hellenic

Page 10: Dynamic IP configuration of terminals in broadcasting networks

G. Gardikis et al. / Computer Networks 52 (2008) 292–302 301

Ministry of Education. The implementation andintegration into a real DTV network is being carriedout in the frame of the EU-funded IST EN-THRONE research project (End-to-end QoSthrough Integrated Management of Content, Net-works and Terminals).

References

[1] ETSI EN 301 192: Digital Video Broadcasting (DVB): DVBSpecification for Data Broadcasting, European Standard,v.1.4.1, ETSI, 2004.

[2] G. Fairhurst, B. Collini-Nocker, Unidirectional LightweightEncapsulation (ULE) for Transmission of IP Datagramsover an MPEG-2 Transport Stream, RFC 4326, December2005.

[3] M. Kornfeld, U. Reimers, DVB-H – the emerging standardfor mobile data communication, EBU Tech. Rev. (January)(2005).

[4] G. Gardikis, G. Kormentzas, G. Xilouris, H. Koumaras, A.Kourtis, Broadband data access over hybrid DVB-T net-works, in: Procedings of the HET-NETs’05, Ilkley, UK,2005, pp. 18–20 (July).

[5] D. Adami, S. Giordano, M. Pagano, R. Secchi, Modeling thebehavior of a DVB-RCS satellite network: An empiricalvalidation, in: Proceedings of the HET-NET’s’05, Ilkley,UK, 2005 (July).

[6] J. Cosmas, T. Itegaki, L.Cruickshank, et al. ConvergedDVB-T and GPRS service scenarios and application pro-duction tools, in: Proceedings of the DTV Workshop,CONFTELE 2003, Aveiro, Portugal, pp. 5–8.

[7] R. Droms, J. Bound, et al., Dynamic Host ConfigurationProtocol for IPv6 (DHCPv6), RFC 3315, July 2003.

[8] R. Droms, Dynamic Host Configuration Protocol, RFC2131, March 1997.

[9] R. Droms, W. Arbaugh (Eds.), Authentication for DHCPMessages, RFC 3118, June 2001.

[10] National Institute of Standards and Technology, NISTNetHome Page, <http://www-x.antd.nist.gov/nistnet/>.

[11] G. Gardikis, G. Kormentzas, A. Kourtis, Dynamic IPconfiguration for interactive digital television terminals, in:IEEE International Conference on Communications 2006/Workshop in IP over Broadcast Networks, Istanbul, Turkey,June 11–16, 2006, pp. 55–59.

Georgios Gardikis received the Diplomain Electrical and Computer Engineeringfrom the National Technical Universityof Athens in 2000 and his Ph.D. fromthe same Institute in 2004. Since then,he has been cooperating with theNTUA Mobile RadiocommunicationsLab and the National Center of Scien-tific Research ‘‘Demokritos’’ in theframe of national and EU-fundedresearch activities, including several IST

projects. At present, he is working as an Associate Researcherin NCSR ‘‘Demokritos’’. He is the technical coordinator of the

IST IMOSAN project, aiming at multi-layer management andtriple-play services provision in DVB-S.2/DVB-RCS networks.He is also collaborating with the University of the Aegean,conducting a national research project regarding experimentalDVB-H deployment. He has gained expertise in the field ofcomposite wireless networking and digital satellite/terrestrialbroadcasting, including DTV picture quality assessment algo-rithms and data service optimisation and multiplexing overDVB platforms. He is also a Member of the IEEE/BroadcastTechnology Society and the Technical Chamber of Greece since2001.

Spiros Orfanos was born in Athens in1981. He holds a bachelor degree inApplied Informatics and Multimediafrom the TEI of Crete and a Msc in DataCommunication Systems from BrunelUniversity. His main interests are in thearea of interactive television, computernetworks, and digital terrestrial TV.He has cooperated as an AssociateResearcher with the ‘‘Pasiphae’’ labora-tory of the TEI of Crete and the Digital

Communications Laboratory of NCSR ‘‘Demokritos’’. He hasactively participated in EU-funded projects ATHENA and

ENTHRONE.

George Kormentzas is Assistant Profes-sor in the University of the Aegean,Department of Information and Com-munication Systems Engineering andresearch associate with the Institute ofInformatics and Telecommunications ofthe Greek National Center for ScientificResearch ‘‘Demokritos’’. He was bornin Athens, Greece on 1973. He receivedthe Diploma in Electrical and ComputerEngineering and the Ph.D. in Computer

Science both from the National Technical University of Athens(NTUA), Greece, in 1995 and 2000, respectively. His research

interests are in the fields of traffic analysis, network control,resource management and quality of service in broadband net-works. He has published extensively in the fields above, ininternational scientific journals, edited books and conferenceproceedings. He is a member of pronounced professional soci-eties, an active reviewer and guest editor for several journalsand conferences and EU-evaluator for Marie Curie Actions. Hehas participated in a number of national and internationalresearch projects at FP5 (i.e., CREDO) and FP6 (i.e.,ENTHRONE) programmes, serving in some instances as theproject’s technical representative for University of Aegean or‘‘Demokritos’’ and/or as WP leader and/or as the project’sTechnical Manager. Currently, he is the technical manager ofFP6-IST-STREP UNITE project and co-chair of the FP6B3GSA cluster.
Page 11: Dynamic IP configuration of terminals in broadcasting networks

302 G. Gardikis et al. / Computer Networks 52 (2008) 292–302

Evangelos Pallis is the director of PASI-PHAE laboratory at CTR of Crete. Hereceived his Ph.D. in Telecommunica-tions in 2002 form the University of EastLondon. Currently, he is working as aProfessor at ATEI of Crete and as aSenior Researcher at the CTRC. Hisresearch interests are in the fields of net-work protocol mechanisms, multimediaapplications, satellite communications,digital video broadcasting (DVB) tech-

nologies, digital interactive television systems and services, andwireless broadband systems and mobile telecommunication tech-

nologies. As a Senior Researcher of CTRC, he has participated ina number of EU-funded research and development projects,including the AC215 ‘‘CRABS’’, IST-2000-26298 ‘‘MAMBO’’,IST-2000-28521 ‘‘SOQUET’’, IST-2001-34692 ‘‘REPOSIT’’, IST-2002-FP6-507637 ‘‘ENTHRONE’’, and as Technical Manager forthe IST-2002-FP6-507312 ‘‘ATHENA’’ project. He has also par-ticipated in a number of National funded projects. He has morethan 40 publications in international journals, conference andworkshop proceedings, in the above areas.

Anastasios Kourtis received his BS degreein Physics in 1978 and his Ph.D. inTelecommunications in 1984, both fromthe University of Athens. Since 1986 heis a Researcher in the Institute of Infor-matics and Telecommunications of theNational Centre for Scientific Research‘‘Demokritos’’, currently ranking asSenior Researcher. His current researchactivities include, digital terrestrialinteractive television, broadband wireless

networks, Perceived Quality of video services, end-to-end QoSand real time bandwidth management in satellite communica-

tions. He is author or co-author of more than 80 scientific pub-lications in international scientific journals, edited books andconference proceedings. He has a leading participation in manyEuropean Union funded research projects in the frame of IST/FP5/FP6 (MAMBO, SOQUET, CREDO, WIN, LIAISON,ENTHRONE). He has also coordinated three European fundedSpecific Targeted Research Projects (REPOSIT, ATHENA,IMOSAN).