harmonizing heterogeneous components in sesame

2
Short Paper: Harmonizing Heterogeneous Components in SeSaMe Luciano Baresi, Sam Guinea and Adnan Shahzada Politecnico di Milano Deep-SE Group - Dipartimento di Elettronica Informazione e Bioingegneria Piazza L. da Vinci, 32 - 20133 Milano, Italy Email: [email protected] Abstract—This paper describes SeSaMe, an RDF-enriched self-adaptive middleware for highly dynamic and autonomic complex systems. Semantic technologies make components of a system mutually interoperable and enable them to exchange information even when they are developed and deployed on different platforms. SeSaMe uses semantics to achieve dynamic component discovery and self-configurability in the network of Things, and to support the communication among heterogeneous Things. The paper exemplifies its contribution for harmonizing heterogeneous components in the context of a smart greenhouse scenario. I. I NTRODUCTION The Internet of Things (IoT) envisions self-configuring and self-organizing systems that can have several thousands of heterogeneous components operating on different platforms, performing diverse sets of tasks, and belonging to different application and administrative domains. For the realization of these systems, we need the support of a middleware that can provide interoperability among components, as well as abstrac- tions for organizing, composing, and managing them [1]. Many researchers have addressed the interoperability of Things by means of middleware infrastructures [1]. Among these, RUNES and SOCRADES support the sophisticated integrations of embedded devices into networks of Things. ASPECS [2] suggests an approach based on holons of in- teracting objects and maintains a hierarchical organization of the components. The Cobra [3] architecture presents a systematic approach for the rapid prototyping of context-aware applications. SETH [4] presents an agent-based hierarchical architecture for the development of smart environments and uses aggregation principles to compose different smart spaces. This paper describes our ongoing efforts on an IoT mid- dleware called SeSaMe (Semantic Self-Adaptive Middleware, [5]). SeSaMe is a group-based middleware that offers a means to deal with large-scale, dynamic, and autonomic systems. In SeSaMe, components can join a system without previously knowing its topological formation. SeSaMe also provides a semantic layer, based on RDF (Resource Description Frame- work, [6]), that allows different system components to ex- change information and communicate using a common model. We believe that semantic models help solve the integration This research was partially funded by the European Commission, Pro- gramme IDEAS-ERC, Project 227977 SMScom. Adnan Shahzada is supported by the Joint Open Lab “S-Cube”, sponsored by Telecom Italia S.p.A. - Innovation division, Milan, Italy. problem, both in terms of component discovery and data exchanges, but we also believe that they can play an important role in coordinating the components. In the context of SeSaMe, the semantic layer has an impact on basic group setup (i.e., on the system’s topology), enables self-configurability, and clarifies how the topology should change, as the system’s set of components and its context of execution evolve. For the semantic layer we leverage Smart-M3 [7], a platform operating on principles of space-based information exchange. II. SMART GREENHOUSE This paper exemplifies the senmatic layer of SeSaMe in the context of a smart greenhouse 1 . It is essential for a greenhouse to be able to utilize its resources optimally, by grouping similar plants together in certain rooms according to their temperature and other physiological needs. The carts in our system vary in terms of their APIs, and in particular in the structure of the messages they can send and receive. The SeSaMe [5] middleware consists of two main layers: a lower-level layer called the Components Layer, and a higher- level layer called the Management Layer. The Components Layer contains SeSaMe components, that is, abstractions that allow different kinds of dynamic Things to participate in the IoT application. On the other hand, the Management Layer contains the SeSaMe Smart Space Managers (SSM). Both layers organize their components under the form of groups. A group comprises multiple components; one of the components is selected to be the group’s supervisor, while the other components are followers. The supervisor component is responsible for managing the components in its group and it is directly connected to a SSM in the Management Layer whereas followers receive directives from the supervisor. The greenhouse is modeled in a way that each room is represented by a group in the Components Layer, and the carts are components that become followers of these groups (see Figure 1(a)). In the Management Layer we have a single Greenhouse SSM that is responsible for supervising the room supervisors. Room supervisors are responsible for collecting data from their followers. Their goal is to identify their follower carts’ needs and report to the SSM if they see any abnormal behavior or data. SeSaMe uses an RDF-based ontological representation of information. The semantic model can be conceptually divided 1 The demo is available online at http://code.google.com/p/sesame-polimi/. 2014 IEEE World Forum on Internet of Things (WF-IoT) 978-1-4799-3459-1/14/$31.00 ©2014 IEEE 197

Upload: malik-muchamad

Post on 17-Dec-2015

7 views

Category:

Documents


0 download

DESCRIPTION

jurnal

TRANSCRIPT

  • Short Paper: Harmonizing HeterogeneousComponents in SeSaMe

    Luciano Baresi, Sam Guinea and Adnan ShahzadaPolitecnico di Milano

    Deep-SE Group - Dipartimento di Elettronica Informazione e BioingegneriaPiazza L. da Vinci, 32 - 20133 Milano, Italy

    Email: [email protected]

    AbstractThis paper describes SeSaMe, an RDF-enrichedself-adaptive middleware for highly dynamic and autonomiccomplex systems. Semantic technologies make components ofa system mutually interoperable and enable them to exchangeinformation even when they are developed and deployed ondifferent platforms. SeSaMe uses semantics to achieve dynamiccomponent discovery and self-congurability in the network ofThings, and to support the communication among heterogeneousThings. The paper exemplies its contribution for harmonizingheterogeneous components in the context of a smart greenhousescenario.

    I. INTRODUCTION

    The Internet of Things (IoT) envisions self-conguring andself-organizing systems that can have several thousands ofheterogeneous components operating on different platforms,performing diverse sets of tasks, and belonging to differentapplication and administrative domains. For the realization ofthese systems, we need the support of a middleware that canprovide interoperability among components, as well as abstrac-tions for organizing, composing, and managing them [1].

    Many researchers have addressed the interoperability ofThings by means of middleware infrastructures [1]. Amongthese, RUNES and SOCRADES support the sophisticatedintegrations of embedded devices into networks of Things.ASPECS [2] suggests an approach based on holons of in-teracting objects and maintains a hierarchical organizationof the components. The Cobra [3] architecture presents asystematic approach for the rapid prototyping of context-awareapplications. SETH [4] presents an agent-based hierarchicalarchitecture for the development of smart environments anduses aggregation principles to compose different smart spaces.

    This paper describes our ongoing efforts on an IoT mid-dleware called SeSaMe (Semantic Self-Adaptive Middleware,[5]). SeSaMe is a group-based middleware that offers a meansto deal with large-scale, dynamic, and autonomic systems. InSeSaMe, components can join a system without previouslyknowing its topological formation. SeSaMe also provides asemantic layer, based on RDF (Resource Description Frame-work, [6]), that allows different system components to ex-change information and communicate using a common model.We believe that semantic models help solve the integration

    This research was partially funded by the European Commission, Pro-gramme IDEAS-ERC, Project 227977 SMScom. Adnan Shahzada is supportedby the Joint Open Lab S-Cube, sponsored by Telecom Italia S.p.A. -Innovation division, Milan, Italy.

    problem, both in terms of component discovery and dataexchanges, but we also believe that they can play an importantrole in coordinating the components. In the context of SeSaMe,the semantic layer has an impact on basic group setup (i.e.,on the systems topology), enables self-congurability, andclaries how the topology should change, as the systems setof components and its context of execution evolve. For thesemantic layer we leverage Smart-M3 [7], a platform operatingon principles of space-based information exchange.

    II. SMART GREENHOUSE

    This paper exemplies the senmatic layer of SeSaMe in thecontext of a smart greenhouse1. It is essential for a greenhouseto be able to utilize its resources optimally, by grouping similarplants together in certain rooms according to their temperatureand other physiological needs. The carts in our system vary interms of their APIs, and in particular in the structure of themessages they can send and receive.

    The SeSaMe [5] middleware consists of two main layers: alower-level layer called the Components Layer, and a higher-level layer called the Management Layer. The ComponentsLayer contains SeSaMe components, that is, abstractions thatallow different kinds of dynamic Things to participate inthe IoT application. On the other hand, the ManagementLayer contains the SeSaMe Smart Space Managers (SSM).Both layers organize their components under the form ofgroups. A group comprises multiple components; one of thecomponents is selected to be the groups supervisor, while theother components are followers. The supervisor component isresponsible for managing the components in its group and it isdirectly connected to a SSM in the Management Layer whereasfollowers receive directives from the supervisor.

    The greenhouse is modeled in a way that each room isrepresented by a group in the Components Layer, and thecarts are components that become followers of these groups(see Figure 1(a)). In the Management Layer we have a singleGreenhouse SSM that is responsible for supervising the roomsupervisors. Room supervisors are responsible for collectingdata from their followers. Their goal is to identify theirfollower carts needs and report to the SSM if they see anyabnormal behavior or data.

    SeSaMe uses an RDF-based ontological representation ofinformation. The semantic model can be conceptually divided

    1The demo is available online at http://code.google.com/p/sesame-polimi/.

    2014 IEEE World Forum on Internet of Things (WF-IoT)

    978-1-4799-3459-1/14/$31.00 2014 IEEE 197

  • Greenhouse SSM

    Room1SV

    Warm Temperature Room

    Cool Temperature Room

    Mild Temperature Room

    Room2SV

    Room3SV

    cart1

    cart4

    cart5cart2

    cart6

    cart8

    cart3cart7

    Components Layer

    Management Layer

    (a) Group Topology

    Knowledge Base

    RDF-Triples

    Management Layer (SSM)

    Components Layer (Things)

    Subscribes to Component Role info

    Inserts Group Info for Component

    Inserts Role Info for Component

    Subscribes for Group Info

    (b) Discovery and self-conguration of compo-nents.

    into a Group Coordination Ontology (GCO), a Message Proto-col Ontology (MPO), and a Domain Concept Ontology (DCO).The GCO denes the generic concepts used by SeSaMefor self-conguration. According to the GCO, every Grouphas multiple Components, and a GroupInfo (Supervisor andFollower Roles). The inter-communication among componentsis carried out according to the MPO that denes the generalmessaging protocol. Not only does it enforce rules aboutcomponents information exchange, it also denes messageexchange types that simplify the interpretation of the messagesthemselves. The DCO denes the vocabulary that is used tostandardize the content of the messages exchanged amongsystem components. In the greenhouse example, DCO denesthe message that contains an instantiation of the carts Healthconcept, or an instantiation of the rooms Status concept.

    SeSaMe offers a mechanism for Things to connect tothe system without having any previous knowledge about thetopology. When different components want to join a systemthey insert their role information (described using the GCO)into the knowledge base in the form of RDF triples. SSMssubscribe to this information (see Figure 1(b)), so they arenotied as soon as a triple is inserted into the knowledge base.They then collaborate to chose a group for the component. Incase of greenhouse, the SSMs will nd a group (room) for theincoming cart that matches the carts role. After choosing agroup for the component, the SSMs insert their decisions intothe knowledge base. The cart is made aware of this input andit connects to the assigned group (room).

    III. EXPERIMENTATION

    We used a context simulator for ubiquitous systems calledSiafu [8] to help demonstrate our SeSaMe middleware. Wehave tested the middleware on two specic scenarios.

    The rst scenario comprises three new ower carts, namedcart1, cart2, and cart3, that want to enter the greenhouse.Cart1 carries owers with high temperature needs, and henceimplements the WarmTemperature Role. Similarly, cart2 andcart3 have CoolTemperature Role and MildTemperature Role,respectively. Group Information about all the potential grouptypes needs to be pre-registered with the Greenhouse SSMto enable autonomic discovery and group assignment. Uponreaching the greenhouse, the carts insert their role informationinto the knowledge base to know which room they shouldgo to. For example, a component with id cart1 with thefollower role CoolTempeatureRole will insert the followingtriple into the knowledge base: SeSaMe : cart1, SeSaMe :FollowerRole, CoolTemperatureRole. The information isreceived by the Greenhouse SSM that nds the appropriate

    rooms for the carts according to their roles (correspondingtemperature needs). In the current example, the SSM allocatescart1 to the Warm Temperature Group, cart2 to the CoolTemperature Group, and cart3 to the Mild Temperature Group.

    The second scenario describes what happens when certainowers become sick. For instance, if some owers on cart3get sick, cart3 uses the DCO to notify the Room3SV throughthe following RDF-triple: SeSaMe : cart3, Greenhouse :Health, Sick, which in turn noties the Greenhouse SSM.In the meanwhile, Room3SV asks cart3 to activate the sickower role and to reconnect to the system, through the SSM,which decides where it should put the sick owers. It canallocate a new room for sick plants, or transform the currentroom into the room for sick owers. In the latter case, therooms supervisor has to change its role, and all other owersneed to leave the room. In this example, SSM decides tomake Mild Temperature Room the room for all sick plants,and changes the role of Room3SV to that of the sick owersupervisor. This will leave all the nodes following the previoussupervisor role in an orphaned state. In SeSaMe, all nodeswhose supervisor fails or leaves the system, automatically tryto reconnect to a SSM and will be allocated to a new room.

    IV. CONCLUSIONS AND FUTURE WORK

    The paper describes how SeSaMe harmonizes heteroge-neous Things by integrating an RDF-based semantic layer.This layer enables component discovery and self-congurationin large-scale, dynamic complex systems. It also supports themanagement of huge networks of Things by dening onto-logical models for forming groups and exchanging messagesamong heterogenous components. Our experiments have onlybeen limited to a simulated smart greenhouse environment,but even in this context, where communication delays arenegligible, we did not experience signicant delays introducedby the semantic layer. In our future work we intend to extendour experimentations to a wider set of systems.

    REFERENCES

    [1] L. Atzori, A. Iera, and G. Morabito, The internet of things: A survey,Computer Networks, vol. 54, no. 15, pp. 2787 2805, 2010.

    [2] M. Cossentino, N. Gaud, V. Hilaire, S. Galland, and A. Koukam, Aspecs:an agent-oriented software process for engineering complex systems,In Autonomous Agents and Multi-Agent Systems, vol. 20, pp. 260304,2010.

    [3] H. Chen, An Intelligent Broker Architecture for Pervasive Context-Aware Systems, Ph.D. dissertation, University of Maryland, BaltimoreCounty, December 2004.

    [4] I. Marsa-Maestre, M. Lopez-Carmona, J. Velasco, and A. Paricio,Mobile devices for personal smart spaces, in Proceedings of 21stInternational Conference on Advanced Information Networking andApplications Workshops, Ontorio, Canada, vol. 2, 2007, pp. 623628.

    [5] L. Baresi, S. Guinea, and A. Shahzada, Sesame: Towards a semantic selfadaptive middleware for smart spaces, in Post-Proceeding of Workshopon Engineering Multi-Agent Systems(EMAS), 2013, pp. 169178.

    [6] Resource description framework, Referenced October 2013. [Online].Available: http://www.w3.org/RDF/

    [7] J. Honkola, H. Laine, R. Brown, and O. Tyrkko, Smart-m3 informationsharing platform, in Proceeding of IEEE Symposium on Computers andCommunications (ISCC), 2010, 2010, pp. 10411046.

    [8] M. Martin and P. Nurmi, A generic large scale simulator for ubiquitouscomputing, in 3rd Annual International Conference on Mobile andUbiquitous Systems - Workshops,., 2006, pp. 13.

    2014 IEEE World Forum on Internet of Things (WF-IoT)

    198