Coordinating allocation of resources for multiple virtual IPTV providers to maximize revenue

826 IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 4, DECEMBER 2011

Coordinating Allocation of Resources for MultipleVirtual IPTV Providers to Maximize Revenue

Sasitharan Balasubramaniam, Member, IEEE, Julien Mineraud, Philip Perry, Brendan Jennings, Member, IEEE,Liam Murphy, Member, IEEE, William Donnelly, and Dmitri Botvich, Member, IEEE

Abstract—Network virtualization is seen by many as a keytechnology to help overcome some of the constraints of the currentInternet architecture and help build a “human centric” FutureInternet. As IPTV gains popularity, creating virtual networksfor IPTV Service Providers can allow them to deploy specificprotocol suites, routing algorithms and resource allocation strate-gies without affecting other IPTV providers that share the sameunderlying infrastructure. However, from the perspective of theunderlying networking infrastructure provider (the “carrier”)virtualization presents new management challenges, in particularhow to efficiently and fairly allocate available resources to multiplevirtual networks. In this paper we describe a framework in whichmanagement systems associated with virtual IPTV provider net-works communicate with the management system of the carrier toprovide coordinated resource allocation. The proposed approachallows policy-based management systems to control a bio-inspiredresource management mechanism, based on species competitionfor biological systems, that a carrier can use to allocate resourcesto competing IPTV providers in a manner that maximizes thecarrier’s revenue. Results of a simulation study are presented,which show that this approach outperforms uncoordinated virtualnetwork management approaches.

Index Terms—Bio-inspired approaches, network management,overlay networks, service management, virtualization.

I. INTRODUCTION

O VER the years, the Internet has continually attractedusers through innovative and diverse services. One

particular service that has gained tremendous popularity inrecent years is IPTV [1]–[7]. For the purposes of this paperwe consider the use of IPTV technology to provide a digitalmultimedia stream over the Internet through triple play, whichincludes data, voice, and video. The video streaming portioncomes in a number of versions, ranging from standard def-inition to high definition for both video on demand as well

Manuscript received March 14, 2011; revised July 15, 2011; accepted July20, 2011. Date of publication September 22, 2011; date of current versionNovember 23, 2011. This work was supported by Science Foundation Irelandvia the “Federated, Autonomic Management of End-to-End CommunicationsServices” strategic research cluster under Grant 08/SRC/I1403, and via the“A Biologically inspired framework supporting network management for theFuture Internet” starting investigator award under Grant 09/SIRG/I1643.

S. Balasubramaniam, J. Mineraud, B. Jennings, W. Donnelly, and D.Botvich are with the TSSG, Waterford Institute of Technology, Waterford,Ireland (e-mail: [email protected]; [email protected]; [email protected];[email protected]; [email protected]).

P. Perry and L. Murphy are with the School of Computer Scienceand Informatics, University College Dublin, Dublin, Ireland (e-mail:[email protected]; [email protected]).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TBC.2011.2164309

as live streaming. The EU’s “Digital Agenda” [8] predictsthat, as the popularity of IPTV services grows, many smallproviders will enter the market to offer either specialist contentor geographically relevant content. To lower the barrier to entryfor such service providers the Digital Agenda expects to seean increased sharing of network resources which, in turn, willrequire low cost management systems to enable such services tobe delivered effectively over these shared networks. Therefore,we will start to witness individual providers creating overlaynetworks on top of the carrier network. The carrier typicallymanages most of the resource management and routing controlon behalf of the IPTV providers. However, in the days beforeIPTV, when traffic patterns were less dynamic, carriers wereable to carry out these management tasks on a medium to longterm basis, at minimal cost, and in a manner that providedadequate Quality-of-Service (QoS) levels.

As the number of IPTV users as well as the diversity of con-tent increases, network operators will see significant increasesin the volume and heterogeneity of traffic carried on their net-works. One hugely significant trend is the growing volume ofpeer-to-peer [5] video streaming traffic, which many predictwill soon become the dominant traffic type on the Internet. Thisis further exacerbated when the number of IPTV providers in-creases. For carriers, these trends mean it will be no longertenable to employ only traditional medium to long term man-agement techniques to support resource requirements of IPTV.They must augment these with techniques for dynamic alloca-tion of resources to each IPTV provider (from here onwardsIPTV service provider will be referred to as an IPTV-P) togetherwith techniques to more frequently reconfigure the networkinginfrastructure (for example to support better routing plans forthe current traffic demands).

Although a number of solutions have been developed forQoS-based routing (for example MPLS path management [9]),these solutions require modifications to underlying routers andare not suitable for frequent route re-configurations. Given this,a de-coupling approach is required to remove the managementburden from the carrier network, whilst offering IPTV providersthe ability to have much more control over how they delivertheir services to their customers. This de-coupling approachis nowadays termed “network virtualization,” with currentvirtualization techniques building on previous work on virtuallocal area networks (VLANs), virtual private networks (VPNs),active/programmable networks and overlay networks.

An illustration of virtual IPTV overlay networks is providedin Fig. 1, which shows two virtual IPTV overlays, one (IPTV 1)spanning more than one carrier network domain and the other(IPTV 2) spanning a single carrier network domain and sharing

0018-9316/$26.00 © 2011 IEEE

BALASUBRAMANIAM et al.: COORDINATING ALLOCATION OF RESOURCES FOR VIRTUAL IPTV PROVIDERS 827

Fig. 1. Virtual IPTV-P overlay networks deployed over two carrier domains.

some of the underlying routers and links with IPTV 1’s virtualoverlay. The creation of virtual IPTV overlay networks allowsIPTV-Ps to deploy management systems that directly controlpath selection between specific source and destination pairs inorder to satisfy particular sets of QoS metric targets relevantfor the particular content. In times of high load the IPTV-P’sdedicated management system can enforce differentiated flowadmission control based on subscription profiles. For example,users subscribing to a “gold” package could be prioritized overusers subscribing to “silver” or “bronze” packages.

Although a number of studies have investigated dual man-agement of a single overlay and its underlying carrier network,investigation into management of multiple overlay networkssharing a common carrier is still in its infancy, and poses anumber of research challenges [10], [11]. In this paper we focuson the following issues:

• How a carrier should allocate its resources to multipleIPTV-P overlays in a manner that:— ensures that the IPTV-Ps have sufficient resources to

meet requirements; and that— maximizes the revenue generated for the carrier?

• How to supply IPTV-Ps with knowledge of the amount ofresources (bandwidth) the carrier is allocating on a givenpath so that they do not over-use resources?

• How can IPTV-Ps request additional resources for pathswhen traffic demands increases?

We describe a framework, based on the policy-based networkmanagement paradigm [12], in which the management systemsassociated with IPTV-P overlays can communicate with themanagement system of the carrier to exchange such infor-mation, thereby collectively providing coordinated resourceallocation for the overall system. We also specify and evaluatean approach based on the Lotka-Volterra species competitionmodel, used in ecology to model the dynamics of resource con-sumption, in which the carrier allocates resources to competingIPTV-Ps in a manner that maximizes its own revenue.

The paper is organized as follows: Section II reviews rel-evant related work on IPTV service provision and overlaynetwork management. Section III discusses IPTV-P overlaynetwork deployments and highlights resource allocation issues.

Section IV provides an outline of our framework for coor-dinated IPTV overlay network management, describing howpolicy based network management systems can be federated tofacilitate exchange of management information, and presentsthe Lotka-Volterra based coordinated resource allocation tech-nique. Section V describes the results of a simulation study thatcompares the proposed coordinated management approach withindependently managed and unmanaged alternatives. Finally,Section VI summarizes the paper and outlines areas for furtherwork.

II. RELATED WORK

In this section we will discuss two key research areas that arerelevant to the solution proposed in this paper, which are IPTVand overlay networks.

A. IPTV

As IPTV moves further towards the mass market, a numberof different research domains have formed. An example of thisis the work that interlinks IPTV to standardization activities out-lined by Maisonneuve et al. [6]. These standardization activitiescover a wide range of aspects including approaches for servicelevel monitoring that combines Quality of Service dimensionsfrom the network and human perception of the delivered ser-vices. Lee et al. [4] presented an architecture for IPTV that hasproved to be useful in enabling the research community to targetoutstanding issues with a common purpose.

Besides standardization and architecture development, abody of research work has also looked at network protocolsand algorithms to support IPTV. Asghar et al. [1] minimizesthe effects of congestion in IPTV networks using a connectionadmission control mechanism in the core combined with areal-time signaling mechanism to provide effective qualitymonitoring. The bandwidth reservation protocol (RSVP) max-imizes the resources usage while protecting the Video QualityExperience (VQE) of accepted users. When congestion occurs,incoming streams will be rejected so that existing streamswill be preserved. The limitations of this mechanism is theunfairness of bandwidth sharing in the event of multi-classtraffic or multiple IPTV-Ps coexisting in the network.

One major challenge that remains is the question of openingthe use of IPTV beyond controlled network domains to bridgeacross administrative boundaries and still maintain control ofservice quality. The approach proposed by Kim et al. [3] is touse a service policy function to enable service quality moni-toring data to adjust the routing and transportation of the IPTVstream. Davy et al. [13] describe two admission control algo-rithms based around estimating the effective bandwidth requiredto satisfy QoS targets for admitted IPTV flows. The approachemploys a revenue-maximizing algorithm that utilizes informa-tion relating to the price, duration and request frequency for dif-ferent content items to make profit-optimal admission controldecisions.

Other techniques have been developed to reduce bandwidthutilization in parts of the network where the popularity ofVideo on Demand (VoD) content is high. For instance, DeVleeschauwer and Laevens [2] enhanced a caching techniqueby migrating copies of the content closer to future users based

828 IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 4, DECEMBER 2011

on the content’s popularity. Lee et al. [5], used knowledge ofnetwork performance to optimize content caching location.

End to end quality adaptation techniques have also been de-veloped so that the server can adapt the bit rate per VoD streamto improve network utilization without unduly sacrificing pic-ture quality [7].

Therefore, as we can see that there are various approachesthat can be adopted to improve the quality of delivered IPTVstreams, whether this may include route adaptations or migra-tion of contents between servers. However, if each IPTV-Padopts different approaches for managing resources, and inparticular if these providers overlay on each other, this couldcreate unbalance resource consumption, which in turn couldaffect the underlying carrier network. The solution presented inthis paper addresses this potential problem, in particular fromperspective of revenue.

B. Overlay Networks

In this section we review a number of studies that have inves-tigated issues relating to management of both single and mul-tiple overlay networks. One of the earliest works on overlaynetworks, was the Resilient Overlay Networks (RON) [14] ap-proach. The objective of RON is to create a fault tolerant net-work that can route around failures with minimal disconnectiontime. The approach targets inter-domain network solutions, re-lying on underlying network routing protocols and providingQoS support for overlay paths.

Koizumi et al. [15] investigate the use of a three-layeredoverlay network, consisting of an overlay layer, a virtualnetwork topology (VNT), and an optical layer. The authorsinvestigate the effect changes at the VNT layer will have onthe optical layer, looking in particular at link utilization. Theproposed solution introduces hysteresis to improve the stabilitybetween the two layers. Stability is improved by minimizingthe interactions between the layers so that the only interactionspermitted have a significant impact on improving performance.Although this is shown to improve link utilization in the opticallayer, the technique would lead to high blocking rates forcustomers, which was not addressed in the study.

Chun et al. [16] evaluated the characteristics of overlay net-work routing using selfish nodes playing competitive construc-tion games. In particular this is applied to multi-domain overlaynetworks, where nodes in each domain want to selfishly behaveto maximize resources for its own domain. Keralapura et al. [17]investigate the co-existence of overlay ISPs network and under-lying IP networks, in particular during network failures. The au-thors point out key challenges for multiple overlay routing. Onesuch challenge is minimizing physical link overload when si-multaneous overlays perform re-routing. At the same time, theauthors also discuss overlays for multiple domains, where routeoscillations in a specific domain can affect and lead to oscilla-tions of entire overlay paths.

Liebnitz et al. [18] propose a self-adaptive overlay networksolution based on bio-inspired techniques. The mechanism isbased on attractor-selection algorithm that considers the pathquality. In the event of quality degradation, the path will auto-matically change to improve the application QoS. Liu et al. [19]investigate the effects of dual routing for overlay and underlay

Fig. 2. IPTV overlay deployment with no management.

networks. The mechanism used is a two-player non-coopera-tive non-zero sum game with two separate objectives for eachlayer, where the overlay objective is to try to minimize delayof its traffic and the native layer’s objective is to minimize net-work costs. Their results indicate that, in some circumstances,the interaction between the dual layers can lead to performancedegradation due to conflicting objectives.

While numerous studies have investigated the stability of dualrouting, including for multiple overlays, to our knowledge nostudies have addressed techniques for managing the allocationof resources to multiple overlays so that carrier resources areutilized efficiently and overlay QoS targets can be met.

III. MULTIPLE ISP OVERLAY DEPLOYMENTS

The basic deployment scenario for IPTV-P overlay net-works is depicted in Fig. 2. The carrier owns and operates theunderlying physical network infrastructure, selling resources(bandwidth between endpoints) to independent IPTV-Ps, whoin turn sell triple play packages to their customers. EachIPTV-P consists of video servers that house VoD content,where these servers are linked to a router of the carrier net-work (e.g. a Video Serving Office (VSO) may be connectedto router 1 of the carrier network, which is represented as

in the overlay of Fig. 2). Each IPTV-P willhave specific delivery points, and this is shown for examplein Fig. 2 as in the overlay layer, which represents egressrouter 6 of the underlying carrier. Therefore, for each of theoverlay nodes, the IPTV-P can perform virtual overlay routingto deliver contents to the various egress points. IndividualIPTV-Ps perform their own route calculation (for exampleusing a link state routing protocol) in accordance with theirmeasured or estimated traffic demand matrix. Since routingcan be performed in the overlay layer, routes for the samesource/destination pair may take different overlay logicallink(s) as well as physical links. For example in Fig. 1, theoverlay path of from overlay node to maytake path , which could map to carrier path

, while for may takeoverlay paths , whichmaps to carrier path . In Fig. 2, IPTV-Ps do

BALASUBRAMANIAM et al.: COORDINATING ALLOCATION OF RESOURCES FOR VIRTUAL IPTV PROVIDERS 829

Fig. 3. Independently managed IPTV-P overlays.

not deploy management systems for their overlay network, thusthey are not able to perform prioritized admission control oftraffic flows in high load conditions. In the paper we refer to thisas the “unmanaged” IPTV overlay approach, or as “solution 1”(“S1”).

Given that IPTV-Ps are not aware that the logical paths oftheir overlay are mapped to carrier paths that are potentiallyshared with traffic from other IPTV-Ps, there is a risk of overutilization of the carrier links [10]. If the carrier has dimen-sioned the network properly this would usually not be a sig-nificant problem, however during link failure or high load con-ditions the operation of carrier and overlay routing algorithmscan easily lead to oscillatory behavior known as route flapping,which significantly degrades customers QoS [17]. From the per-spective of the IPTV-P, the fluctuating availability of carrier re-sources can be mitigated through deployment of resource allo-cation techniques at the overlay layer. For example the deploy-ment of a Policy-based Management System (PBM) [12], as de-picted in Fig. 3 can allow each IPTV-P to deploy their own poli-cies, where these policies may incorporate flow admission con-trol which ensures that when demand outstrips the availabilityof resources, flows associated with more valuable customersare prioritized over flows associated with less valued customers[20]. For example in Fig. 3, can use PBM to deter-mine which overlay route a the traffic of a gold customer shouldtake as opposed to traffic from a silver customers. In the eventthat load on the overlay links start to increase, then policies mayindicate that traffic flows from silver customers should be throt-tled in order to ensure that gold customers’ traffic continues toreceive acceptable QoS. In the case of , a differentoverlay route may be taken for their gold customers.

In this paper we refer to this as the “independently managed”solution, or a “solution 2” (“S2”). A clear shortfall with this ap-proach is that IPTV-Ps are limited to managing their own over-lays only, they do not have visibility of the degree to whichthe underlying carrier resources they use are shared with otherIPTV-Ps. On the other hand the carrier does not have the abilityto adaptively allocate resources to individual IPTV-Ps so thatthe they can use its links without overloading them.

Fig. 4. Coordinated multi-overlay management system for overlay and carrier.

IV. COORDINATED MULTI-IPTV OVERLAY MANAGEMENT

In this paper, we seek to provide a solution for coordi-nated management of multiple ITPV-P overlays [21], [22].The coordinated management proposed here is between theIPTV-Ps and the carrier, whereby the carrier allows IPTV-Psto dynamically change the level of resources leased from thecarrier. This offers the carrier the opportunity to maximizeits revenue, whilst simultaneously ensuring that the collectivebehavior of the overlay networks does not lead to sub-optimalperformance for any of the IPTV-Ps. To achieve this there mustbe communication between the management systems of theIPTV-Ps and the carrier, as depicted in Fig. 4. This communi-cation would take the form of requests from an IPTV-P to thecarrier for additional resources on a specified overlay path tomeet changing customer demands for that IPTV-P’s services.Following such requests the carrier may reallocate its resourcesbetween the IPTV-Ps it supports and will then inform them ofthe new allocations. This must be done within the constraintsof the physical bandwidth available:

(1)

where is the capacity of the physical link betweennodes and , while is the capacity of the overlaylink for between nodes and .

We assume that carrier and IPTV-P management systems usethe policy-based network management paradigm, in which thebehavioral rules for network management are separated fromthe code that realizes the functionality of given network devices.Policies are typically formulated as event-condition-actions tu-ples with the semantics of “on event(s), if condition(s), do ac-tion(s).” These rules can be specified by network administra-tors and deployed directly onto network devices (via configu-rations applied typically through command-line interfaces), orare maintained by independent Policy Decision Points, whichspecify actions to be enforced when notified of events by the net-work devices. Examples of the application of the policy based

830 IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 4, DECEMBER 2011

Fig. 5. PBM support for multiple IPTV overlay networks.

management paradigm in single administrative domains are dis-cussed by Zhuang et al. [23] and Agrawal et al. [24].

In our framework we create a federated policy based man-agement system, in which the actions resulting from policiestriggered in the IPTV management system are interpreted asevents which trigger policy evaluations in the carrier manage-ment system and vice versa. Policies allow system administra-tors to easily configure management system behavior; for ex-ample, an IPTV-P policy may contain a condition clause thatindicates that the action of requesting additional resources fromthe carrier should only be executed if a certain threshold per-centage of traffic flows for “silver” customers have been rejectedby an admission control process within a given elapsed time in-terval. As depicted in Fig. 5 policies can be separated into “in-ternal” policies and “federated” policies, where the latter con-trol interaction between the management systems. The internalpolicies can be tailored to suit each IPTV-P’s objective (e.g. calladmission), and this could be linked to federated policies (e.g.request for resource of a certain quantity). The federated policiesgovern when requests for additional resources are forwarded tothe carrier.

In this paper we assume that all policies are specified in asingle language and are related to a single information model,such that the syntax and semantics of the policy actions ex-changed between the systems are interpreted correctly. How-ever, incorporating policies that have different syntax and se-mantics for each IPTV-P is also possible, but semantic map-pings between the systems will be required. This topic is out ofthe scope of this paper, but further discussion of how this canbe achieved can be found in van der Meer et al. [25]. The nextsub-sections will present details of our proposed solution. Ini-tially we describe the process of resource allocation between thecarrier path and the overlay virtual links. This will be followedby description of internal operations within the IPTV-P basedon the resources allocated from the carrier, and the last sub-sec-tion will present the Lotka-Volterra competition model that isused to distribute resources between the physical path in the un-derlying network, and the overlay links.

A. Carrier Path Resource Mapping to Overlay Virtual Links

As described in the introduction, one of the objectives of oursolution, is to allow the carrier to allocate resources to eachITPV-P, so that each ITPV-P will be able to perform operations

Fig. 6. Algorithm for internal IPTV-P Operation.

(e.g. re-routing at the overlay layer) without affecting the per-formance of other IPTV-Ps. In order to allow this level of sta-bility, the carrier must have an accurate picture of each IPTV-Presource requirements, and accurately distribute the resourcesfrom the physical path in the underlying network to the vir-tual overlay links of the IPTV-Ps. We will explain this con-cept through an example presented in Fig. 7(a). In this example,there are two IPTV-P overlay networks sharing resources from asingle carrier. Initially, the carrier determines the optimum pathbetween different pairs of nodes in the carrier network. In thisexample, the optimum path of ,and and are direct links between the nodes.This is then followed by each IPTV-P requesting resources be-tween its overlay nodes from the carrier. For example if wehave a particular link with 1 Gbps of capacity and two possibleIPTV-Ps competing for resources, the bandwidth may be allo-cated at a ratio of 20:80 at some given instant of time. There-fore, for both and , the carrier willmap to , to , and to

(this is triggered by ) shown in Fig. 7(a).Once the resources have been allocated, each overlay networkdetermines the optimum routes at the overlay layer, as shownin Fig. 7(b) (dashed lines). Based on the resources allocated tothe virtual links of each IPTV-P, found the optimalpath for to be the direct virtual link between nodesand , while for the optimal path for is

.In this sub-section, we have only presented the process of

mapping and allocating resources between the paths in the car-rier layer and the overlay virtual paths. The mechanism of di-viding the resources for each IPTV-P’s overlay virtual link isbased on the Lotka-Volterra model, which will be presented inSection IV-C.

B. IPTV-P Internal Operations

Once resources are allocated to the IPTV-Ps, each IPTV-Pwill be able to manage the resources according to its revenueobjectives and users’ service requirements. As described in theprevious sub-section, each IPTV-Ps will be able to performrouting at the overlay layer. Therefore, if congestion is encoun-tered on an overlay virtual link, the IPTV-P will be able toperform re-routing based on the resources allocated to it by thecarrier, without having to be concerned that such operation may

BALASUBRAMANIAM et al.: COORDINATING ALLOCATION OF RESOURCES FOR VIRTUAL IPTV PROVIDERS 831

Fig. 7. Example of coordinated management of multiple overlay IPTV-Ps.

affect the performance of other IPTV-Ps. The decision to per-form re-routing rather than purchase more resources could bebased on some knowledge of the traffic behavior. The IPTV-Pcould evaluate the traffic pattern, and find that the resourcepreviously purchased from the carrier is sufficient and doesnot require to bid for further resources. This also enables thecarrier to maintain a certain degree of stability in the system,where the carrier is likely to configure its policies so that itwill not immediately trigger a re-allocation of resources everytime it receives a request from an IPTV-P. Instead it couldtrigger re-allocations based on all requests received within agiven interval—for example over the course of a day or a week.At the same time, if an IPTV-P has gold, silver, and bronzesubscription packages (or similar) for their customers, priori-tized admission control can be made for different subscriptionpackages when resource become scarce. An algorithm to showthe internal operation of the IPTV-P is presented in Fig. 6.

Initially, each checks the current availablecapacity of its overlay links everyseconds. In the event that an existing overlay path capacity

between overlay nodes and is below a threshold, the IPTV-P will recalculate the path for .

Each will calculate the load on the overlay net-work, and when the load exceeds the threshold, the

will select incoming traffic based on probabilityvalues for each subscriber package class (lines 6–9). Fig. 7(c)presents an example of this process. The figure shows that eachITPV-P has performed a re-route for , which is trig-gered by the internal of each IPTV-P. In this ex-ample, triggers a re-route when the threshold of theoverlay path approaches a certain threshold of the total overlaypath capacity.

C. Lotka Volterra Model

Section IV-A presented the allocation of resources from thecarrier path to the overlay virtual links. This section will de-scribe the process of distributing the resources to each IPTV-Pdepending on their demand. This mechanism is through theLotka-Volterra competitive symbiosis model [26]. We willfirst describe the Lotka-Volterra model, and describe how thismodel is used to coordinate resource management between thedifferent IPTV-Ps.

The Lotka-Volterra model was developed to model the com-petition between multiple species competing for a fixed set ofresources. Our application is inspired by the work of Kodama etal. [27], who applied it to management of TCP congestion. TheLotka-Volterra competition model is represented as:

(2)

where represents the type of species , represents thegrowth rate, represents the ratio of competition betweenspecies and , whilst represents the carrying capacity ofthe environment. The species could represent a total aggre-gate aggressiveness from the remaining species in the competi-tion. Therefore, the model is not only limited to a single species,where Eq. (2) could represent the competition between species

and the total aggressiveness of remaining species, representedas . The model represents the consumption of resources by dif-ferent species depending on their aggressiveness. For speciesthat are highly aggressive, the consumption of resources willbe high compared with less aggressive species. However, thespecies are still able to co-exist symbiotically. Fig. 8 illustratesan example of how resources are being consumed for differentcompetition (values of ) between different species. In Fig. 8(a),

832 IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 4, DECEMBER 2011

Fig. 8. Effect of varying competition in the Lotka-Volterra modes.

when all species have low competition values, each consumesvery similar quantities of resources. Similarly this could be seenin Fig. 8(b), where all species have medium aggressiveness.Fig. 8(c) shows a slightly mixed strategy, where low aggressivespecies will consume less compared to species with high aggres-siveness. Fig. 8(d) shows this more clearly with wider range inaggressiveness between the different species. While the modelshows transients behaviors, we only consider the final resourceconsumption once the model approaches steady-state.

In our application, the competition between the IPTV-Ps isanalogous to the environment of competing species. IPTV-Psmust continually evolve to improve sustainability and meet con-tinually changing environments (e.g. user demand). This is anal-ogous to species that must survive and evolve by consuming re-sources in order to maintain survivability in face of any changesfrom the environment. We, therefore, express the competitionmodels for virtual overlay link as:

(3)

where represents the amount of capacitythe carrier allocates to from capacity of carrierpath for source-destination , the sparecapacity of , and represents the competitiveness of

. We assume a constant adaptation rate for each

Fig. 9. Algorithm for carrier resource management.

IPTV-P. The is dependent on a number of factors, which in-cludes the demand from users and the amount of revenuegenerated by the IPTV-P.

Fig. 9 presents the algorithm for the carrier resource manage-ment. Initially, the carrier calculates the optimal path betweenthe nodes in the underlying network. The carrier is responsiblefor recalculating paths for all source and destination pairs whena certain path is loaded. Based on the overlay link for

, the carrier will distribute resources from the phys-ical path depending on the competition valueof . All resource distribution is synchronously per-formed every seconds. Each IPTV-P is also notconfined to their original competition status of resources fromthe carrier, where each IPTV-P can change their competitionvalue to obtain new resources. The carrier will

BALASUBRAMANIAM et al.: COORDINATING ALLOCATION OF RESOURCES FOR VIRTUAL IPTV PROVIDERS 833

TABLE ISIMULATION PARAMETERS

charge the IPTV-P a certain amount whenthe IPTV-P changes the .

Through this approach, IPTV-Ps who wish to selfishly holdback unused resources will have to pay a higher price whenthe demand is high, leading to lower revenue. This solutionprovides scalability to large numbers of IPTV-Ps and alsoallows resources to be dynamically requested when the demandchanges in accordance with IPTV-P policies. At the same time,the framework affords carriers the opportunity to minimizeoverloading physical links that are shared by overlay links ofdifferent IPTV-Ps.

Fig. 7(d) shows the federated operation of resource request.In this example triggers , which eval-uates the historical traffic pattern and determines that extraresources is required from the carrier. Therefore,determines the amount of resources required and invokes

which subscribes the resource from the carrier.The carrier notes that is not fully utilizing itsresources, and since is offering a higher bid(through ), decides to recall the unused resourcefrom and allocate this to .

V. SIMULATION STUDY

The simulation is divided to two sections: Section V-Apresents performance evaluation between three managementapproaches for the multiple overlay networks (unmanagedIPTV-P overlay (S1), independently managed IPTV-P (S2),and Coordinated Multi IPTV-P Overlay Management system(S3); and Section V-B investigates how adaptive changes ofIPTV-Ps competition affect the revenue of other IPTV-Ps. Thetraffic parameters used for the simulation are created syntheti-cally to simulate extreme conditions that IPTV-Ps of the futuremay encounter.

Fig. 10. Carrier revenue.

Fig. 11. Moderate intensity IPTV-Ps’ revenue.

A. Static Competition

The parameters used for the simulation are shown in Table I.We simulate three types of overlay IPTV-Ps, based on thenumber of requests generated by users in each, where thelowest traffic rate is categorized as moderate, medium trafficrate is high intensity, and very high traffic rate is very highintensity; this is also reflected in the ratio of incoming traffic,where out of 80 requests/s, 8% are allocated to each moderateIPTV-P, 12% to each high intensity overlay IPTV-P, and16% to the very high intensity overlay IPTV-P. The selectionpolicy parameters are based on the traffic rate intensity thatis measured on each overlay network. In the event that theload reaches 65% of the capacity allocated by the carrier, theIPTV-P will begin admission control of incoming request basedon probability . Also provided are the parameters for thecompetition model, where we allocated a range of load for eachparameter range (e.g. light load as 0.8).

The simulator is implemented as a discrete time event sim-ulation written in Java. The results shown in Figs. 10–12 indi-cate that the carrier gains the largest revenue using S3. Figs. 13and 14 show the very high intensity overlay IPTV-P gains thehighest revenue with S3 compared to S2 and S1 (for Euros/secearned). This is mainly because the IPTV-P has been allocated

834 IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 4, DECEMBER 2011

Fig. 12. High intensity IPTV-Ps’ revenue.

Fig. 13. Very high intensity IPTV-Ps’ revenue.

Fig. 14. Average throughput of high intensity IPTV-Ps.

resources based on its requirement and therefore is able to max-imize the resources subscribed to suit its requirements. This isalso reflected in the throughput shown in Fig. 15, where S3 re-sulted in the highest throughput, followed by S2 and S1. At ap-proximately time , S1 has higher throughput than S2and S3, and this is because of the selection process used by S2and S3, which leads to limiting the amount of traffic admittedinto the network. Since S1 indiscriminately allowed all trafficinto the network early, this leads to higher loading of the overlay

Fig. 15. Average throughput of very high intensity IPTV-Ps.

Fig. 16. Path capacity comparison between solution 1, 2 and 3.

network. However, after , S3 was able to admit moretraffic, and this is due to the higher capacity allocated by thecarrier.

The advantage of the competition model is also illustrated inFig. 16, which shows the spare capacity of overlay path for aspecific source-destination pair of very high intensity overlayIPTV-P. The graph shows the reduction in spare capacity asthe path is loaded with respect to time. The load is the numberof incoming requests for a specific overlay source and destina-tion pair. The fluctuations indicate new path calculation events,when the old path is loaded to a certain threshold (in this case90%). As shown in the figure, the use of competition model pro-vides the resource that suits the intensity of the IPTV-P’s trafficdemand. In this case, the very high intensity overlay IPTV-P hadthe highest share, compared to the other IPTV-Ps at the initialstage of the simulation.

Fig. 12 shows that one of the high intensity overlay IPTV-Ps(all behave similarly) has a similar behavior to the very high in-tensity IPTV-P, although the revenue gained is not as high. Theaverage throughput is shown in Fig. 14 and, as expected, the S3throughput is less than S1 at , again because of theselection process. However, as the network slowly gets loaded,S3 is able to gain higher revenue compared to S2 and S1. How-ever, this behavior is not present in moderate overlay IPTV-Ps

BALASUBRAMANIAM et al.: COORDINATING ALLOCATION OF RESOURCES FOR VIRTUAL IPTV PROVIDERS 835

(Fig. 11), where the competition does not result in higher rev-enue compared to even distribution of resources provided by S2.

B. Adaptive Competition

This section will evaluate the performance of overlayIPTV-Ps when they adaptively change their competitive status.The scenario for the simulation is based on changes in trafficpattern, and the ability for the IPTV-P to change its com-petitiveness. The performance evaluation will compare thesituation if the IPTV-P had adaptive competitive status com-pared to constant status (e.g. moderate or very high intensity),to measure the gain the adaptability provides. The change intraffic pattern will be performed based on time zones, wherea certain request rate stays constant during the time zone, butchanges as it transitions between the time zones. For scenarioA there are three time zones, while scenario B has 6 timezones. Scenario A will evaluate the effect a single IPTV-P(IPTV-P 7) with adaptive competitive status will have on otherIPTV-Ps which have constant competitive status. Scenario Bwill evaluate the effect of the system when there are multipleoverlay IPTV-Ps adaptively changing their competitive status.The internal policies of the IPTV-Ps are as follows:

• : When the measured load of the IPTV-Preaches a certain threshold, trigger prioritized call admis-sion (call admission is based on probability presented inTable I).

• : When the measured path load of theIPTV-P reaches a certain threshold, if traffic pattern isstable, then re-route.

The federated policy for the IPTV-Ps is as follow:• : In the event of changes in traffic de-

mand, determine new resources required and submit newvalue to the carrier.

As described earlier, the carrier is responsible for deter-mining the most appropriate resources for each IPTV-P usingthe Lotka-Volterra model. The methodology is based on dy-namically changing prices for the resources, depending on thecompetitiveness and demand of resources from all IPTV-Ps.Therefore, during peak periods, resources will be chargedhigher as the demand increases. The internal policy for thecarrier is as follows:

• : If underlying network path load reachesa certain threshold, perform re-route calculation.

The federated policy for the carrier is as follows:• : Every , recalculate the re-

sources for each ISP depending on their latestvalue.

1) Scenario A: This scenario uses the same topology andpayload configuration detailed in Table I , but with average ser-vice time of 20 s and request arrival rates for the individualIPTV-Ps as shown in Table II.

The simulation is separated in three time zones, where eachzone shows the traffic rate (requests/second) for each IPTV-P.

The scenario starts in zone 1 with IPTV-P1 having very high intensity status, IPTV-Ps 2,3, and 4 havinghigh intensity, and the other 6 IPTV-Ps having moderate inten-sity. At zone 2 , IPTV-P 7 changes its status to

TABLE IINUMBER OF REQUESTS/SECOND FOR EACH IPTV-P (SCENARIO A)

Fig. 17. Progressive revenue comparison between adaptive status to constantmoderate and very high intensity status.

very high intensity; it then changes its status back to moderatein zone 3 .

Fig. 18(g) shows the gain of IPTV-P 7 when it is in adaptivestatus compared to constant very high intensity and moderateintensity. Therefore, this illustrates the improvement adaptivestatus has over constant status when the traffic behavior changesbetween zones. Fig. 17 presents the progressive revenue with re-spect to time for IPTV-P 7. As shown in the figure, the total rev-enue of the adaptive status gives the highest revenue comparedto constant moderate and very high intensity. The performanceof this adaptive regime is compared to two static regimes, onewhere IPTV-P 7 adopts a moderate intensity state and the otherwhere IPTV-P 7 adopts a very high intensity state for the dura-tion of the simulation. Results of these simulations are shown inFig 17. They all show a loss of revenue compared to when theIPTV-P 7 adopts a static moderate state.

Fig. 18 also presents the results of revenue comparison be-tween the adaptive, constant moderate, and constant very highintensity status of IPTV-P 7, and its effect on the performanceof other IPTV-Ps. The green shaded areas in Fig. 18 shows thegain each IPTV-P makes when IPTV-P 7 has a adaptive com-petitiveness compared to constant very high intensity, and thered area shows the gain each IPTV-P makes when IPTV-P 7has an adaptive competitiveness compared to constant moderatestatus. For example in Fig. 18(a), IPTV-P 1 (with constant veryhigh intensity status) shows that in zone 1 and 3, it is able toreceive higher revenue if IPTV-P 7 had an adaptive status. Al-though zone 2 would yield higher revenue if IPTV-P 7 had aconstant moderate intensity, the overall revenue gain is higher

836 IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 4, DECEMBER 2011

Fig. 18. Revenue for scenario A (y-axis is /s). (a) Revenue for IPTV-P 1. (b) Revenue for IPTV-P 2. (c) Revenue for IPTV-P 3. (d) Revenue for IPTV-P 4. (e)Revenue for IPTV-P 5. (f) Revenue for IPTV-P 6. (g) Revenue for IPTV-P 7. (h) Revenue for IPTV-P 8. (i) Revenue for IPTV-P 9. (j) Revenue for IPTV-P 10.

if IPTV-P 7 is adaptive. Fig. 18(a)–(d), shows the impact ofIPTV-P 7’s adaptive status to the IPTV-P’s 2, 3, and 4’s highintensity status. Similar to IPTV-P 1’s performance, the overallrevenue is high when IPTV-P7 has an adaptive status, eventhough zone 2 showed a higher revenue gain if IPTV-P 7 hada moderate status. Fig. 18(e) and (f) and Fig. 18(h)–(j) showsthat if IPTV-P 7 had an adaptive status, the moderate IPTV-Psbenefit with higher revenue overall only in zones 1 and 2.

All results in Fig. 18 show reduced revenue if IPTV-P 7adopts a very high intensity state as the red line is alwayslower than the blue. The adaptive regime reduces this loss andincreases the revenue for IPTV-P 7.

2) Scenario B: Unlike the previous scenario, where onlya single overlay IPTV-P changed its status, this scenario con-siders a number of overlays changing their status through 6 timezones. The number of requests for each zone for each IPTV-Pis shown in Table III. The scenario for the multiple IPTV-Pschanging their status is based on IPTV-Ps 1-5 increasing their

traffic demand at different rates (IPTV-P1 increases at thehighest rate, followed by IPTV-P2, and IPTV-P3, 4, and 5increases at a lower rate). IPTV-Ps 6-10 had a reasonably con-stant traffic pattern. Fig. 17 shows the revenue performance ofall the IPTV-Ps., whilst Fig. 19 shows the revenue performanceof individual IPTV-Ps. As shown in the figure, IPTV-P 1 whichhad the highest aggressiveness, yields the highest revenue withthe adaptive status, compared to constant high intensity andmoderate intensity.

As shown in the figures, IPTV-P 2-5 also benefit from adap-tive status as their traffic rate increases from zone to zone. Thisis especially shown towards the latter zones (e.g. zone 5-6).IPTV-Ps 6-10 demonstrate with a relatively constant revenuerate due to the relatively constant traffic pattern throughout allthe time zones (as shown in Table III). The initial time zones,show that the IPTV-Ps benefit from the adaptive status (althoughthe performance fluctuates tremendously). The most significantresult is that the resources have been shared in a way the yields

BALASUBRAMANIAM et al.: COORDINATING ALLOCATION OF RESOURCES FOR VIRTUAL IPTV PROVIDERS 837

TABLE IIINUMBER OF REQUESTS/SECOND FOR EACH IPTV-P (SCENARIO B)

Fig. 19. Revenue for scenario B (y-axis is /s). (a) Revenue for IPTV-P 1. (b) Revenue for IPTV-P 2. (c) Revenue for IPTV-P 3. (d) Revenue for IPTV-P 4. (e)Revenue for IPTV-P 5. (f) Revenue for IPTV-P 6. (g) Revenue for IPTV-P 7. (h) Revenue for IPTV-P 8. (i) Revenue for IPTV-P 9. (j) Revenue for IPTV-P 10.

good revenues by using the policy driven species competitionmodel.

VI. CONCLUSION

As the popularity of IPTV increases, the underlying com-munications infrastructure of the Internet will see an increasein virtual IPTV-Ps sharing resources over the carrier net-works. This paper has presented a framework, based on the

policy-based network management paradigm to manage coor-dinated multi-IPTV-P overlay networks. The proposed solutionallows the carrier to fairly distribute resources to individualIPTV-Ps based on their demand and competitiveness withrespect to other IPTV-Ps. Through the environment of compet-itiveness, the system shows improved revenue for the carrierand for the IPTV-Ps with high and very high traffic intensity.At the same time, the coordinated resource distribution, allows

838 IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 4, DECEMBER 2011

each IPTV-P to efficiently manage resources without affectingthe performance of other IPTV-Ps. A simulation study wasconducted to compare the superiority of the coordinated man-agement systems with unmanaged and independently managedoverlay IPTV-Ps. At the same time, the study also evaluatedthe effects of IPTV-Ps changing their competitive status, andthe impact this has on other IPTV-Ps.

The results show that the proposed framework can use abio-inspired species competition approach to implement policydriven resource management. This approach shows a robustmethod to enable network virtualization and federated networkmanagement. While the solution presented in this paper is onlylimited to a single carrier domain, the solution can definitely beextended to multiple domains through federated managementsystems as outlined in [28].

REFERENCES

[1] J. Asghar, F. Le Faucheur, and I. Hood, “Preserving video quality inIPTV networks,” IEEE Trans. Broadcast., vol. 55, no. 2, pp. 386–395,Jun. 2009.

[2] D. De Vleeschauwer and K. Laevens, “Performance of caching algo-rithms for IPTV on-demand services,” IEEE Trans. Broadcast., vol. 55,no. 2, pp. 491–501, Jun. 2009.

[3] T. Kim, J. Kim, and M. Hahn, “Policy-based service overlay IPTVframework for open strategy,” in IEEE Int. Conf. Netw. Infrastructureand Digital Content, Sep. 2010, pp. 962–966.

[4] G. M. Lee, C. S. Lee, W. S. Rhee, and J. K. Choi, “Functional architec-ture for NGN-based personalized IPTV services,” IEEE Trans. Broad-cast., vol. 55, no. 2, pp. 329–342, Jun. 2009.

[5] S.-B. Lee, G.-M. Muntean, and A. F. Smeaton, “Performance-awarereplication of distributed pre-recorded IPTV content,” IEEE Trans.Broadcast., vol. 55, no. 2, pp. 516–526, Jun. 2009.

[6] J. Maisonneuve, M. Deschanel, J. Heiles, W. Li, H. Liu, R. Sharpe, andY. Wu, “An overview of IPTV standards development,” IEEE Trans.Broadcast., vol. 55, no. 2, pp. 315–328, Jun. 2009.

[7] G.-M. Muntean, P. Perry, and L. Murphy, “A comparison-based studyof quality-oriented video on demand,” IEEE Trans. Broadcast., vol. 53,no. 1, pp. 92–102, Mar. 2007.

[8] “Communication from the commission to the European parliament, thecouncil, the European economic and social committee and the com-mittee of the regions: A digital agenda for Europe,” European Com-mission, Aug. 2010.

[9] C. Scoglio, T. Anjali, J. de Oliveira, I. Akyildiz, and G. UhI, “TEAM:A traffic engineering automated manager for diffserv-based MPLS net-works,” IEEE Commun. Mag., vol. 42, no. 10, pp. 134–145, Oct. 2004.

[10] M. Ammar and S. Seetharaman, “Routing in multiple layer: Chal-lenges and opportunities,” in Proc. Workshop Internet Routing Evol.Des. (WIRED), Oct. 2006.

[11] N. Chowdhury and R. Boutaba, “Network virtualization: state of theart and research challenges,” IEEE Commun. Mag., vol. 47, no. 7, pp.20–26, Jul. 2009.

[12] J. Strassner, Policy-Based Network Management: Solutions for the NextGeneration. San Mateo, CA: Morgan Kaufmann, 2003.

[13] A. Davy, D. Botvich, and B. Jennings, “Revenue optimized IPTV ad-mission control using empirical effective bandwidth estimation,” IEEETrans. Broadcast., vol. 54, no. 3, pp. 599–611, Sep. 2008.

[14] D. G. Andersen, H. Balakrishnan, M. F. Kaashoek, and R. Morris, “Re-silient overlay networks,” in Proc. 18th ACM Symp. Oper. Syst. Prin-ciples (SOSP), 2001, pp. 131–145.

[15] Y. Koizumi, T. Miyamura, S. Arakawa, E. Oki, K. Shiomoto, andM. Murata, “On the stability of virtual network topology control foroverlay routing services,” in Proc. 4th Int. Conf. Broadband Commun.,Netw. Syst. (BROADNETS), Sep. 2007, pp. 810–819.

[16] B.-G. Chun, R. Fonseca, I. Stoica, and J. Kubiatowicz, “Characterizingselfishly constructed overlay routing networks,” in Proc. 23rd Annu.Joint Conf. IEEE Comp. Commun. Soc. (INFOCOM), Mar. 2004, pp.1329–1339.

[17] R. Keralapura, N. Taft, C. nee Chuah, and G. Iannaccone, “Can ISPstake the heat from overlay networks?,” in Proc. ACM SIGCOMM Work-shop Hot Topics Netw. (HotNets), 2004.

[18] K. Leibnitz, N. Wakamiya, and M. Murata, “Biologically inspired self-adaptive multi-path routing in overlay networks,” ACM Commun. Lett.,vol. 49, pp. 62–67, Mar. 2006.

[19] Y. Liu, H. Zhang, W. Gong, and D. Towsley, “On the interactionbetween overlay routing and underlay routing,” in Proc. 24th Annu.Joint Conf. IEEE Comp. Commun. Soc. (INFOCOM), Mar. 2005, pp.2543–2553.

[20] S. van der Meer, A. Davy, S. Davy, R. Carroll, B. Jennings, and J.Strassner, “Autonomic networking: Prototype implementation of thepolicy continuum,” in Proc. 1st Int. Workshop Broadband Converg.Netw. (BcN), Apr. 2006, pp. 1–10.

[21] Y. Zhu and M. Ammar, “Algorithms for assigning substrate networkresources to virtual network components,” in Proc. 5th IEEE Int. Conf.Comput. Commun. (INFOCOM), Apr. 2006, pp. 1–12.

[22] J. Fan and M. H. Ammar, “Dynamic topology configuration in serviceoverlay networks: A study of reconfiguration policies,” in Proc. 25thIEEE Int. Conf. Comput. Commun. (INFOCOM), Apr. 2006, pp. 1–12.

[23] W. Zhuang, Y.-S. Gan, K.-J. Loh, and K.-C. Chua, “Policy-based QoS-management architecture in an integrated UMTS and WLAN environ-ment,” IEEE Commun. Mag., vol. 41, no. 11, pp. 118–125, Nov. 2003.

[24] D. Agrawal, K.-W. Lee, and J. Lobo, “Policy-based management ofnetworked computing systems,” IEEE Commun. Mag., vol. 43, no. 10,pp. 69–75, Oct. 2005.

[25] S. van der Meer, B. Jennings, D. O’Sullivan, D. Lewis, and N. Agoul-mine, “An initial investigation into semantic models for policy mo-bility,” in Proc. 12th Workshop HP OpenView Univ. Assoc. (HPOVUA),2005, pp. 211–225.

[26] J. D. Murray, Mathematical Biology: I. An Introduction (Interdiscipli-nary Applied Mathematics) (Pt. 1), 3rd ed. New York: Springer, Dec.2007.

[27] M. Kodama, G. Hasegawa, and M. Murata, “Implementation experi-ments of TCP symbiosis: bio-inspired mechanisms for Internet con-gestion control,” in Commun. Quality Rel., Apr. 2008.

[28] B. Jennings, K. C. Feeney, R. Brennan, S. Balasubramaniam, D.Botvich, and S. van der Meer, Autonomic Network ManagementPrinciples: From Concepts to Applications. New York: Elsevier,2010, pp. 101–118.

Sasitharan Balasubramaniam (M’05) receivedthe Bachelor’s degree in electrical and electronicengineering and Ph.D. degree from the Universityof Queensland in 1998 and 2005, respectively, andthe Masters degree in computer and communicationengineering in 1999 from Queensland University ofTechnology.

He joined the Telecommunication Software andSystems Group (TSSG), Waterford Institute ofTechnology (WIT), Ireland right after completionof his Ph.D. He is currently the manager for the

Bio-inspired Network research unit at the TSSG. Sasitharan has worked in anumber of Irish funded projects (e.g. Science Foundation Ireland, PRTLI) andEU projects. His research interests include Bio-inspired networks, as well asmolecular communications.

Julien Mineraud received the Bachelor and Master’sdegrees in computing from University Denis Diderot,Paris VII, France in 2004 and 2006, respectively. Heis currently a Ph.D. student at the TelecommunicationSoftware and Systems Group, Waterford Institute ofTechnology.

His research interests include bio-inspired auto-nomic network management, in particular, routingand resources management.

Philip Perry received the B.Eng. degree in electricaland electronic engineering from Strathclyde Univer-sity, Glasgow, Scotland in 1987, the M.Sc. in RF andmicrowave engineering from University of Bradford,England, in 1988, and the Ph.D. from University Col-lege Dublin, Ireland in 1998.

He has held a number of academic and research po-sitions in both University College Dublin and DublinCity University covering both Computer Science andElectronic Engineering schools. He has also workedwith and for a number of companies, most notably

BALASUBRAMANIAM et al.: COORDINATING ALLOCATION OF RESOURCES FOR VIRTUAL IPTV PROVIDERS 839

Lucent Technologies and Ericsson Systems Expertise. Currently he is a consul-tant engineer, and a senior research fellow in both University College Dublin andDublin City University. He has authored 80 papers and is a member of IEEE.

Brendan Jennings (M’06) is a Senior ResearchFellow with the Telecommunications Software &Systems Group, Waterford Institute of Technology,Ireland. His research interests include autonomicnetwork management, management of federations,charging and billing, and performance management.He regularly serves on the technical program com-mittees of a number of network management relatedconferences and workshops and is currently secre-tary of the IEEE Communications Society TechnicalCommittee on Network Operations and Management

(CNOM). More information available at http://brendanjennings.net.

Liam Murphy (S’86–M’91) received the B.E.in electrical engineering from University CollegeDublin in 1985, the M.Sc. and Ph.D. degrees inelectrical engineering and computer sciences fromthe University of California, Berkeley in 1988 and1992, respectively.

He is currently an Associate Professor in Com-puter Science at University College Dublin, wherehe is Director of the Performance Engineering Lab-oratory. He has published over 130 refereed journaland conference papers on various topics, including

multimedia transmissions, dynamic and adaptive resource allocation algorithmsin computer/communication networks, and software development. His currentresearch projects involve mobile and wireless systems, and computer networkconvergence issues. He has established and maintained a number of meaningfulcollaborations with academic and industry partners, both within his disciplineand across disciplinary boundaries. Prof. Murphy is a Member of the IEEE

(Communications, Broadcasting, and Computer societies) and a Fellow of theIrish Computer Society.

William Donnelly received the Bachelor degree inphysics and mathematics from Dublin Institute ofTechnology in 1984, and the Ph.D. in experimentalphysics from University College Dublin in 1989.

He is currently the Head of Research and Inno-vation at Waterford Institute of Technology, Irelandas well as director and founder of the Telecommu-nications Software and Systems Group. As directorof the TSSG, he has overseen over 100 funded Irishand European research projects incorporating basicresearch (HEA PRTLI program and Science Founda-

tion Ireland), applied research (EU IST FP5 FP6, FP7) and commercializationof ICT results (Enterprise Ireland, EU eTEN). His research interests are in theareas of bio-inspired networks and management solutions for next-generationInternet-based electronic media.

Dmitri Botvich (M’07) received the Bachelor’s andPh.D. degrees in mathematics from Moscow StateUniversity, Faculty of Mechanics and Mathematics,Russia, in 1980 and 1984, respectively.

He is currently the Chief Scientist of the Scien-tific and Technical Board at the TelecommunicationSoftware and Systems Group, Waterford Institute ofTechnology (Ireland). He currently leads the PRTLIFutureComm project at the TSSG, and has coordi-nated and worked in a number of EU and ScienceFoundation Ireland projects. He has published over

one hundred papers in conferences and journals, and currently supervises 7Ph.D. students. His research interests include bio-inspired autonomic networkmanagement, security, trust management, wireless networking, queuing theory,optimization methods, and mathematical physics.