arXiv:1105.2458v3 [cs.NI] 28 Jul 2014 sapral hn ratbe C aho hc a be specific can (i.e., which of profile each technical capabilities) PC, communication specific and tablet computational such a a hardware, devices, or by mobile phone characterized several portable concurrently, a use, as may Specifical is devices. user scenarios heterogeneous new a very these adopt of optimiz may feature and users key guarantee that A of to services. order organization on” in “always and required management are among devices and the new these diffusion, cooperation for mass solutions the their the effective the of to foster use owing ubiquitous However, that and devices. pervasive scenarios the through new individuals to rise th of) (parts communication. of hav and we implementations computation results subsystem. preliminary and experimental from some computation introduce obtained effecti of we the forms illustrate approach, to our th ubiquitous order show in and subsystem Finally, we communication. novel a following enables as the implemented that be cooperati in can and addition, framework that integration In the technologies. effectively Wha these support available. protocols, mechanisms, can already (i.e., that framework are a ecosystem is missing digital ne still technologies a and solutions construct t the of to Throughout deployed most peer-to- dissemination. that conveniently show DOs, data we paper, and be among includin sharing can and resource resources, must favor mechanisms DO use DO own interaction the for the its peer ecosystem Inside of of the organization consumption. management of internal power feature of rest The peculiar issues the DOs. ” of address to peer “digital set offer its a a can from his/he as it by with of resources characterized equipped though and subsystem user be a can human (DO), devices, a distri i.e. novel mobile context, organis of own ecosystem”, this development “digital the “digital In so-called foster services. a can of that term out below), (see we constructed enables users what ecosystem human an of by construction GPS used PCs, commonly the tablet players) telephones, media portable gators, (e.g., devices mobile http://dx.doi.org/10.1109/IWCMC.2011.5982863. optn nDgtlEoytm:Dsg susadChallen 978- Communicati and ISBN Workshop. 9538-2. Wireless Platforms Issues Computing International and Mana Communication Design Emergency Panzier 7th (IWCMC-2011), conference Computing Fabio Ecosystems: the Mobile of Ghini, Digital Proceedings Vittorio in Ferretti, Computing Stefano D’Angelo, 0 h rlfrto fmbl,htrgnosdvcsgives devices heterogeneous mobile, of proliferation The Terms Index Abstract h ulse eso fti ae saalbeat available is paper this of version publisher The I hspprw ru httesto wireless, of set the that argue we paper this —In oieCmuigi iia Ecosystems: Digital in Computing Mobile Pe-oPe;Ntok;Modeling Networks; —Peer-to-Peer; .I I. NTRODUCTION areeDAgl,SeaoFret,Vtoi hn,Fabi Ghini, Vittorio Ferretti, Stefano D’Angelo, Gabriele einIse n Challenges and Issues Design eateto optrSine nvriyo Bologna of University Science, Computer of Department { laect s Gabriele as: cite Please dneo fret hn,panzieri ghini, sferrett, gdangelo, eesof veness .Mobile i. services) n and ons gement: 1-4577- oon,Italy Bologna, buted nof on navi- eded ms” and this is t ges. ly, se at at o e g e r s , meddi ag ope system. complex large a in embedded hn nefc fte“et vial ewr sacellula a cellular is a network to available access “best” give the if may interface gateway phone the criteria communica such (e.g., available that to “best network” “gateway” or according example, special interface for a level”, output “battery through the via pass select other should will each DO with a communicate informatio outside to Secondly, device technologies. able different wireless be the use short-range should firstly the DO summary, coordinate a In and in resources. optimize share to those can transparent and must that of (fully use they users) mechanisms human to ecosystem, adaptive their home DOs with their enable endowed in to be available (and order if resources In manual or, the user. out through other human carried only each operations the occur with configuration interact may complicated) not usually interaction do of do, DO de capable they technological different same made often, the of Most be within resources. number should their DOs rationally a using overcome. that be requires limitations issues, mobility enable to order in hosts, it them. DOs among the ecosyst cooperation the to by transparently deployed m not be itself, heterogeneities policies are these t that opportunity ecosystem practice, require an In same provide they cooperation. the rather, promote problem; populating a as heterogene DOs regarded possible the principle, (self-)organiz in among their embeds Hence, facilitates it interaction. that DOs and environment of digit attain an community good the they with a provide the as Therefore, should long entities. enhance ecosystem those as can among services, of coordination communication, ubiquitous entities sharing of information, cooperating development the data, among for as computing) mechanisms (intended system own optimization (P2P) its resources with Peer-to-Peer that it on shows providing Literature by contribut resources. the ecosystem it within unused addition, that interaction In its be maintaining embedded. of have is to can forms may it which own DO DOs in its Each ecosystem interacting ecosystem”. and of “digital objectives a own interact community as may a of which thought (DO) DOs; organism” peer “digital with a as of thought raiigaslddgtleoytm ul wr fall of aware fully ecosystem, digital solid a Organizing be can user mobile each scenarios, these within Thus, } @cs.unibo.it Panzieri o sent n vices ation ities tion em by ay as es al o s s r network, or a laptop or PDA Wi-Fi interface, if the “best” 2) The DO needs to be equipped with a software module, available network is an available Wi-Fi network; note that running on a given device of the DO, which acts as “best” here may mean “with lowest latency”, or “cheaper” a gateway that exploits multi-criteria, adaptive decision depending on specific human user QoS requirements). Thirdly, schemes to understand which is the best network in- and finally, computation should be performed by the device terface to use to send/receive information to/from the with the highest computational capability. outside world. 3) Smart P2P schemes must be employed among DOs to Once individuals are organized as entities able to optimally share data/resources. These schemes should take into move and act within the digital ecosystem, they must be consideration: i) social aspects (e.g. users might want to enabled to interact with the external environment in the best share resources only with their friends); ii) trust, security possible way. From this perspective, how to opportunistically and privacy issues; iii) smart discovery strategies to exploit all available technological solutions is a key issue. In identify neighbors owning resource/data of interest; iv) order to enjoy “always on” services, users should be able to tit-for-tat schemes [18]. move transparently from one network to another without losing 4) While interactions among digital are purely their active connections. local, such a local organization should reflect a wider After individuals have been endowed with the needed view of the global overlay, so as to structure the net- technologies for adaptive living in the digital ecosystem, it is work based on some desired topology [8]. Indeed, the natural to envision several forms of cooperation among users, structure of the network is of paramount importance to in order to enhance and globally optimize the overall resource guarantee effective data dissemination [9]. use for given user-specific needs. In particular, users can This paper is organized as follows. In the next Section we freely organize themselves into ad-hoc networks (MANETs) introduce the background related to the topics addressed in and share resources. For example, a user might decide to this paper. Section III discusses our objectives. Section IV share his laptop’s Wi-Fi network interface with a neighbor introduces the principal issues to be addressed in the design of user; the devices in this latter user accessing that interface an effective digital ecosystem and illustrate some performance will communicate with it via Bluetooth or infrared channels. results we have obtained from a partial implementation of This technique can be extended so as to allow the interaction an ecosystem. Finally, Section V provides some concluding between devices for handling computations on remote hosts. remarks. The substantial difference between the digital ecosystem model of interaction to be developed and the traditional II. BACKGROUND ones is that each DO in an ecosystem is to be regarded The massive and pervasive use of personal computers as as a possible computational and communication resource, well as other devices such as smart cellphones, PDAs, new not simply as a node in a network that has to forward generation household appliances and the like, combined with messages towards their final destination (a feature which will the widespread availability of Internet connections on one nevertheless be guaranteed in the digital ecosystem). side and wireless networks on the other, raise new important research issues, commonly grouped together under the caption In this paper, we discuss issues of design of a digital ecosys- of “”. A large community of researchers tem to be deployed in pervasive environments. The ecosystem is investigating these issues, and a host of technologies for can be developed and optimized based on a new paradigm ubiquitous computing has indeed been developed (for the sake we propose; this paradigm is in turn based on self-organizing, of conciseness, we shall not examine them in this paper). For opportunistic networking, and the P2P communication model. instance, both the project “seamless computing” [5] proposed The approach we propose can be summarized in the following by Microsoft in 2003, and Jini, which is a part of the Java tech- four steps: nology developed by Sun Microsystems, have addressed issues 1) Resources available at a given mobile user must be in- of ubiquitous computing. Both these initiatives stimulated an tegrated using autoconfiguration strategies. In particular, initial strong interest. However, they were rapidly forgotten by the Personal Area Network (PAN) should be configured the research and software developers communities as they did so that all the data generated/collected at a given device not introduce any solution or product worth of mention. In the can be made available to other devices of the same context of large scale computing architectures, metacomputing user, ready to be used for any type of computation. [11] has been another attempt to propose a new paradigm The configuration should be accomplished taking into for the organization of large computer networks. Although consideration power consumption issues. Such a dy- metacomputing has achieved rather promising results, a main namic integration converts a set of owned devices into problem with it, which is still open, is the lack of integration a real DO that can exploit all the technological features between the global infrastructure and the users’ devices. available to the user. Furthermore, in the dynamic con- figuration strategies, it will be necessary to consider user A. Interaction And Sharing Mechanisms level policies such as the cost issues related to service Several interaction mechanisms exist for the sharing and ex- usage (e.g. mobile and broadband connectivity). change of resources and information. Seti@Home is probably one of the first system that implements the wireless channel with the primary user in an opportunistic on different hosts. In addition, several techniques exist today manner. Other proposals focus on the decomposition of for the distributed resource utilization; these techniques are ad-hoc networks into small overlapping clusters instead; these usually referred to as “cloud computing” or “ad-hoc data are exploited to build a tree-like network, with the aim of processing” [16]. prolonging the network lifetime through a more efficient As to the resources discovery, several P2P alternatives exist energy utilization. such as publish-subscribe systems, DHT (Distributed Hash Table) based solutions, application level multicast [7]. Finally, the most recent research efforts in hybrid networks As to data dissemination, overlay nets are useful to deliver concentrate on the integration of autonomous wireless net- messages among a set of nodes in a distributed system. works (WLANs, WPANs, cellular) [19]. These efforts put However, when one tries to adopt such an approach in a mobile the focus on topology discovery and maintenance, cross-layer scenario, it must create a mobile ad-hoc network (MANET). optimization and scalability [10]. In fact, a MANET lets nodes communicate among themselves even when no access points are available. The problem is III. DESIDERATA that when an overlay network among mobile nodes is to be The idea of identifying novel interaction paradigms created, additional requirements emerge, such as those related which allow an optimal organization of resources both at to the limited computational and communication capabilities global and local levels, thus overcoming the metacomputing of mobile nodes, their battery capacities, the fact that nodes limitations mentioned earlier, might appear rather ambitious may join/leave the network dynamically. It is well-known and complicated. However, most of the technologies needed that the task of building a P2P network over a MANET to achieve these target paradigms are already available; is not an easy task [13]. Indeed, the proximity of nodes yet, they do not fully integrate with each other. These often suggests to set them as neighbors in the communication technologies include opportunistic, self-organizing networks overlay, while, from an application point of view, they do not and P2P interaction mechanisms. In addition, trust, security actually need to communicate and exchange messages. Thus, and privacy concerns are mandatory issues that need to be in order to optimize the communication among nodes, it is addressed. worth considering both the topology of nodes in real space, as well as the interaction requirements among these nodes at Thus, the main objective to pursue is a novel distributed the application level. architecture that enables the development of ubiquitous appli- cations, where all devices available to a user are dynamically B. Use of Multiple Networks and adaptively configured depending on: i) the devices them- An important feature to be provided is the ability to allow selves, ii) the environment in which they are deployed, and iii) a mobile node, having multiple wireless network interfaces, the other users (and their characteristics) and the interactions to change network points of attachment (handover) without they have. The architecture must thus provide configuration disrupting existing connections. An example of solution to protocols for the intra Personal Area Networking (PAN), support such a host mobility is Mobile IPv6. This solution, to automatically organize all devices belonging to a single implemented at the network layer, requires support from human being seen as a DO. At the same time, it is necessary both network infrastructure and end-nodes. In contrast, to identify and mechanisms for the simultaneous Proxy Mobile IPv6 relies on network infrastructure support and adaptive use of different communication networks in an only. The Host Identity Protocol requires support only from opportunistic fashion. In fact, the overall goal is to optimize the end-nodes. The Terminal Mobility Support Protocol ensures interactions across all DOs populating the ecosystem. Once the IP address transparency by resorting to external SIP (Session organism is able to make the best use of its devices within the Initiation Protocol) proxies. MMUSE works at the application digital ecosystem, it becomes necessary to develop protocols layer and supports vertical handovers by using SIP proxies, for the organization and interaction between different DOs. placed at the edge of the access networks [4]. In [21] a In other terms, mechanisms for the efficient aggregation and scheme is proposed that uses cross-layer information to sharing are needed. accomplish seamless handovers. A similar, more general solution is the IEEE 802.21 standard which defines a IV. DESIGN ISSUES framework (Media Independent Handover services) for the In this section we introduce the principal design issues that exchange of information among different layers, regardless of need to be addressed in order to meet our desiderata. These the low-level technologies. issues are discussed below in isolation.

Research works on ad-hoc networks investigate a wide A. Optimizing the Digital Organism set of issues, and in particular, radio spectrum management, Full interaction among all computing resources (and in hierarchical organization, location-based routing, nodes general all hardware) belonging to the user is needed. This addressing and energy consumption. In this scenario, requires to optimize the use of networks that are available to cognitive radio technology allows a secondary user to share the user’s devices. For these interactions it is natural to adopt 1 Configuration Require: myId device id Require: uId user id Require: myProfile profile of myId 1: for each network interface i do 2: devs[i] = discoveryDeviceList(myId,uId) 3: send(devs[i], myProfile) 4: profiles[i] = receiveFromAll() 5: end for 6: 7: createInternalOverlay(devs, profiles) 8: coordinator = electCoordinator(profiles) 9: gateway = electGateway(profiles)

short-range communication technologies, such as Bluetooth, Fig. 1. Seamless connectivity obtained using multiple communication technologies infrared, ZigBee, etc. This can be achieved through heuristic methods, but also (and perhaps most importantly) by means of decisional procedures based on several criteria. Key metrics are, of course, the computational capability of each device, B. Optimizing the Digital Organism's Interactions within the the available bandwidth and the number of simultaneously Digital Ecosystem interacting devices. The DOs must be provided with a set of protocols to interact with peer DOs within their home ecosystem. These protocols The goal is to model each DO as a computational would allow a DO to opportunistically and dynamically adapt environment and to find the best configuration for it. the interaction with its peer DOs, by selecting the best com- Specifically, based on the computational capacities of each munication protocol among the available ones (e.g. Always device within the PAN, the battery levels, and the available Best Connected, ABC) [12]. The identification of the best network interfaces, devices must be configured so as to available network may be based on multiple criteria such as: identify a primary computation entity, a primary gateway to i) bandwidth, ii) connection cost, iii) battery consumption, iv) send/receive data from the outside world, secondary network of maintaining the connection active while moving, interfaces (e.g. short range ones) to allow communications in order to minimize the hand-offs. Any of these criteria, alone with neighbor organisms. or together with the others, can be used for assessing the best available network at any time. A much more challenging target Algorithm 1 reports the general scheme that devices, is to provide techniques and methods for the simultaneous and forming the DO, should execute during the configuration opportunistic use of all available multiple networks [10]. phase of the PAN. In essence, the idea is that devices must By redirecting traffic across multiple network interfaces discover all other devices belonging to the same user uId. one may suitably balance each network’s load. This approach Since devices communicate through their available network has the fundamental advantage of permitting communication interfaces, each device must concurrently perform such a even when one of the various operating networks suddenly discovery process. In fact, a given device, having two (or falls. To achieve all this, it is important to employ suitable more) network interfaces, might act as a relay to interconnect communication solutions which, for example, make use other devices embodying different networking technologies. of proxies and cross-layer mechanisms in order to firstly Hence, a first goal of this configuration phase is to create an distributing the information flow over multiple communication intra-PAN communication overlay on heterogeneous networks. channels and next re-aggregating the whole flow together again [10]. This seems the only way for making the use An important aspect is that all nodes must distribute all of multiple network interfaces transparent at the application their technical details among each other, in order to enable a level. In fact, the above process should be fully transparent proper configuration of the DO. Different alternatives exist to to the final users (with the exception of security and privacy characterize profiles of devices, such as, for instance, CC/PP issues, which may require an explicit user intervention in [1]. Such information is exploited to identify the coordinator, order to be dealt with). i.e. the device that acts as the resource manager of the DO. To accomplish this task, all devices’ profiles must be distributed Figure 1 illustrates how effective the use of multiple net- among the whole device set internal to the DO, and some works can be in order to provide seamless communications. distributed algorithm must be executed to elect the coordinator. The chart refers to the use of a (really implemented) cross- A similar approach must be employed to identify which device layer architecture that uses a proxy-based system to offer is to act as the primary gateway that manages communications continuity in the communication of a given mobile node with with the outside world. a remote proxy [10]. The mobile node chart was equipped with two different network interfaces, i.e, a Wi-Fi card and nodes that are close to each other in terms of social features. a UMTS card. During the path performed by the mobile node, a temporary unavailability of the Wi-Fi network was Dissemination strategies must be adopted to distribute experienced. Nevertheless, the mobile node was able to auto- messages. In a wide and heterogeneous network such as that matically switch to the UMTS network, without interrupting under consideration, it is important to obtain smart algorithms the end-to-end communication with the remote node. This that allow to broadcast messages among the network nodes. was possible using a proxy which hides the handover from a Such a functionality would allow to implement techniques for network to another (and hence, also the change of IP address the discovery of nodes, resources, paths to transmit messages for the mobile node). In the figure, the experienced latencies towards a destination. It is important to notice that, due to the for packets transmissions are reported; needless to say, those high dynamic of the network, such schemes might be based related to packets delivered through UMTS are higher than on unstructured P2P solutions [8], [14]. Gossip dissemination those sent via Wi-Fi. Nevertheless, a continuous communica- strategies are thus extremely important in this contexts [15]. tion was guaranteed to the application [10]. This represents just a (real) example that demonstrates how viable solutions can be Figure 2 demonstrates that, when adequately tuned, gossip employed to optimize the interactions of the DOs, if it is able algorithms allow to disseminate messages in a network. Three to automatically employ all the resources which are available examples of gossip protocols are considered: i) a message (in this case, two different types of networks), transparently arriving at a node is forwarded to all the node neighbors to its human user. with a given probability (Conditional Broadcast); ii) a message arriving at a node is forwarded to each neighbor with a C. Optimizing the Digital Ecosystem given probability (Fixed Probability); iii) a message arriving The next step is to organize the DOs as a network of at a node is forwarded to a fixed number of (randomly interacting nodes. Many approaches to achieve this goal selected) neighbors (Fixed Fanout) [9]. The charts report the are possible such as overlay or mesh networks, or in average coverage of message dissemination (i.e. the percentage certain settings mechanisms based on not fully decentralized of nodes that receive a specific broadcast) and the average approaches [2]. The main aim is at modeling nodes’ delay (i.e. number of hops), depending on the dissemination interactions while taking full advantage of the typical probability, the time to live (ttl) exploited for disseminated paradigms in terms of complex networks and adaptive messages, and the size of cache set at nodes. systems. In this perspective, the theory of complex networks might help to identify the best topology to organize the users’ D. Optimizing the Computation interactions. For instance, in certain scenarios, the use of Once created the infrastructure for communication and random networks with (almost) uniform nodes’ degrees could interaction, one may handle all computationally demanding be employed, while in other contexts “small worlds” and queries in an adaptive manner, thereby implementing a true scale-invariant networks could be better choices [6]. cloud computing system over mobile ad-hoc technologies [3]. Users may require computations to be performed on remote Moreover, nodes need not only be able to organize and/or unknown hosts, trusted servers, or hosts belonging themselves at random, but also and most importantly to the same user domain. To this end, one firstly needs according to local strategies based on nodes’ targets, to identify/discover those hosts which are able to handle resources and services (both demanded and offered). Other the submitted requests, and next identify the best way to parameters for deciding whether to create a link or not include submit it. In this framework, necessary properties are privacy, the human identities behind the nodes. More specifically, accountability and non repudiability. In non-mobile scenarios, given the privacy concerns raised by some data to be routed these issues have been studied at length; within grid and cloud through the network, it is natural that connections will be computing, ad-hoc processing and P2P computing systems, preferably created among trustworthy DOs. there are many well established useful results [17], [20]. In contrast, in the context of mobile computing, these issues lack This task is easy to handle if the network configuration universally accepted solutions and are still under investigation. is left to the user. Conversely, if the goal is to obtain a (semi) automatic configuration mechanism, then it becomes V. CONCLUSIONS crucial (and hard) to implement an approach based on In this paper, we have discussed a methodology to optimize the grouping of nodes (e.g. using metrics similar to those the interactions of mobile users (in the paper referred as digital employed in social applications). In such a case, network organisms) into dynamic and heterogeneous environments configuration becomes a non-negligible problem, in that one (termed digital ecosystem). The idea is to optimize the use and has to simultaneously take into account: i) nodes’ topology interaction of the devices available to each digital organism, in the environment, ii) the type of network to be created, iii) through dynamic and adaptive configuration strategies (opti- the type of services that nodes are actually seeking/offering, mization of the PAN). Specifically, the interactions within a iv) node heterogeneity in terms of communication and DO and among DOs can be optimized using both the available computation capabilities, v) the clustering/proximity among communication infrastructures and P2P ad-hoc interactions. Gossip protocols evaluation: coverage Gossip protocols evaluation: coverage 100 Cond. Broadcast, pb=50%, ttl=6, cache=256 100 Fixed Prob., v=50%, ttl=6, cache=256 Fixed Fanout., fanout=5, ttl=6, cache=256 80 95

90 60

85 40 80 Percentual coverage Percentual coverage

20 75 Cond. Broadcast, pb=80%, ttl=6, cache=256 Fixed Prob., v=80%, ttl=6, cache=256 Cond. Broadcast, pb=100%, ttl=6, cache=256 Fixed Prob., v=100%, ttl=6, cache=256 0 70 50 100 150 200 250 300 350 400 450 500 50 100 150 200 250 300 350 400 450 500 Number of nodes Number of nodes

Gossip protocols evaluation: delay Gossip protocols evaluation: delay 10 10 Cond. Broadcast, pb=50%, ttl=6, cache=256 Cond. Broadcast, pb=80%, ttl=6, cache=256 Fixed Prob., v=50%, ttl=6, cache=256 Fixed Prob., v=80%, ttl=6, cache=256 Fixed Fanout., fanout=5, ttl=6, cache=256 Cond. Broadcast, pb=100%, ttl=6, cache=256 Fixed Prob., v=100%, ttl=6, cache=256 8 8

6 6

4 4 Delay (number of hops) Delay (number of hops)

2 2

0 0 50 100 150 200 250 300 350 400 450 500 50 100 150 200 250 300 350 400 450 500 Number of nodes Number of nodes Fig. 2. Performances of different gossip schemes. Coverages and number of hops for different of dissemination

Our next step will be to complete the implementation of our [9] S. Ferretti and G. D’Angelo. Multiplayer online games over scale-free approach. We expect to be able to report on our results in the networks: a viable solution? In Proc. of the International Workshop on DIstributed SImulation and Online gaming (DISIO 2010) - Conference near future. on Simulation Tools and Techniques (SIMUTools 2010). ICST, 2010. [10] S. Ferretti and V. Ghini. A web 2.0, location-based architecture for a REFERENCES seamless discovery of points of interests. In Proceedings of the 2009 Fifth Advanced International Conference on Telecommunications, pages 226–231, Washington, DC, USA, 2009. IEEE Computer Society. [1] CC/PP Information Page, http://www.w3.org/Mobile/CCPP/, December 2010. [11] I. Foster and C. Kesselman. Globus: a Metacomputing Infrastructure [2] S. Androutsellis-Theotokis and D. Spinellis. A survey of peer-to-peer Toolkit. International Journal of High Performance Computing Appli- content distribution technologies. ACM Comput. Surv., 36(4):335–371, cations, 11(2):115–128, June 1997. 2004. [12] V. Ghini, S. Ferretti, and F. Panzieri. A Strategy for Best Access Point [3] O. Babaoglu, M. Jelasity, A.-M. Kermarrec, A. Montresor, and M. van Selection. In Proc. of 3rd IEEE/IFIP Wireless Days 2010 Conference. Steen. Managing clouds: a case for a fresh look at large unreliable IEEE, October 2010. dynamic networks. SIGOPS Oper. Syst. Rev., 40:9–13, July 2006. [13] I. Gruber, R. Schollmeier, and W. Kellerer. Peer-to-peer communication [4] M. Bonola, S. Salsano, and A. Polidoro. Upmt: universal per-application in mobile ad hoc networks. Ad Hoc & Sensor Wireless Networks, 2(3), mobility management using tunnels. In Proceedings of the 28th IEEE 2006. conference on Global telecommunications, GLOBECOM’09, pages [14] H. Guclu and M. Yuksel. Limited scale-free overlay topologies for 2811–2818, Piscataway, NJ, USA, 2009. IEEE Press. unstructured peer-to-peer networks. IEEE Trans. Parallel Distrib. Syst., [5] M. Chalmers and I. Maccoll. Seamful and seamless design in ubiquitous 20(5):667–679, 2009. computing. In Proceedings of Workshop At the Crossroads: The [15] M. Haridasan and R. van Renesse. Gossip-based distribution estimation Interaction of HCI and Systems Issues in UbiComp. 2003, 2003. in peer-to-peer networks. In 7th International Workshop on Peer-to-Peer [6] G. D’Angelo and S. Ferretti. Simulation of scale-free networks. In Systems (IPTPS '08), 2008. Simutools '09: Proc. of the 2nd International Conference on Simulation [16] B. Hayes. Cloud computing. Commun. ACM, 51:9–11, July 2008. Tools and Techniques, pages 1–10, ICST, Brussels, Belgium, 2009. ICST. [17] S. Ma, X. Sun, and Z. Guo. A resource discovery mechanism integrating [7] P. T. Eugster, P. A. Felber, R. Guerraoui, and A.-M. Kermarrec. The p2p and grid. In and Information Technology many faces of publish/subscribe. ACM Comput. Surv., 35:114–131, June (ICCSIT), 2010 3rd IEEE International Conference on, volume 7, pages 2003. 336 –339, 2010. [8] S. Ferretti. Modeling Faulty, Unstructured P2P Overlays. In Proc. of [18] D. S. Menasch´e, L. Massouli´e, and D. Towsley. Reciprocity and barter the 19th International Conference on Computer Communications and in peer-to-peer systems. In Proceedings of the 29th conference on Infor- Networks (ICCCN 2010). IEEE, August 2010. mation communications, INFOCOM’10, pages 1505–1513, Piscataway, NJ, USA, 2010. IEEE Press. [19] A. Mitseva, N. Prasad, P. Todorova, R. Aguero, A. Garcia Armada, C. Panayioto, A. Timm-Giel, and L. Maccari. CRUISE research activities towards Ubiquitous Intelligent Sensing Environments. IEEE Wireless Communications Magazine, 15(4):52–60, August 2008. [20] S. Reyes, C. Mu˜noz Caro, A. Ni˜no, R. Sirvent, and R. M. Badia. Monitoring and steering grid applications with grid superscalar. Future Gener. Comput. Syst., 26:645–653, April 2010. [21] K. Tsukamoto, T. Yamaguchi, S. Kashihara, and Y. Oie. Implementation design of handover management scheme for efficient multimodal com- munication. In Proceedings of the 2008 IEEE International Conference on Wireless & Mobile Computing, Networking & Communication, pages 684–689, Washington, DC, USA, 2008. IEEE Computer Society.