You are on page 1of 8

2013 Eighth International Conference on P2P, Parallel, Grid, Cloud and Internet Computing

Towards the Cloud of Things


Sensing and Actuation as a Service, a key enabler for a new Cloud
paradigm

Salvatore Distefano Giovanni Merlino, Antonio Puliato


Dipartimento di Elettronica, Informazione e Bioingegneria Dipartimento di Ingeneria
Politecnico di Milano Universita di Messina
Piazza L. Da Vinci 32, 20133 Milano, Italy Contrada di Dio, S. Agata, 98166 Messina, Italy
Email: salvatore.distefano@polimi.it Email: {gmerlino,apuliafito}@unime.it

AbstractCloud computing has emerged as one of the most about content: there we have crowdsourcing at every corner,
effective paradigms to enable cheap, elastic and user-friendly and the impact of such avenues is never appreciated enough,
access to computing, storage and networking resources. How- witness the undisputed success and overwhelming rise of
ever, a key drawback of current Cloud solutions lies in the lack
of avenues for interaction with the physical world. This work everything social. By the same token, we believe ordinary
aims at contributing the design of an architecture for pervasive users can and should be considered (potential) providers as
ICT infrastructure, where new generation services leverage well, not only of content, but of infrastructure (and metadata
the surrounding environment, collecting data and actuating about it) as well. Therefore we are going to outline an
control strategies. In accordance with such vision computing, effective, abstract model for Clouds, where infrastructure
storage, networking and sensing t as a whole in an all-round
encompassing Cloud-inspired provisioning model, paving the includes I/O in the form of sensors and actuators, provided
way for innovative and value-added services, to be implemented also by volunteering end-users.
by bridging Clouds with the Internet of Things. Heterogeneous On a wider scope and potentially deeper impact, this
resources should be aggregated and abstracted according to paper also introduces things as infrastructure for Cloud-
tailored thing-like semantics, thus enabling a Things as a like exploitation. In this sense we aim to provide a holistic
Service paradigm and eventually disclosing what we call a
Cloud of Things (CoT). approach that allows to combine any kind of resource in
one pervasive infrastructure, disclosing new use cases where
Keywords-Cloud computing; volunteer contribution; re- value-added services mixing context-awareness, augmented
sources mashup; Sensors and Actuators as a Service; Things
as a Service; ontology. reality and social aspects can be provided through the Cloud.
A. Trends in computing
I. I NTRODUCTION
An impressive number of ever more powerful and inte-
Everybody is talking about Clouds nowadays, and this one grated devices (smartphones, household appliances, etc.) join
could easily be the tech mantra of the decade, and beyond, the Internet every year, and the global trafc volume (data,
if early success stories are any indication of things to come. voice, multimedia, etc.) is on the rise accordingly, foreshad-
Yet, for all the hype and interest, both enabling technologies owing a Web of smart devices, or things in the Internet
and use cases appear to be quite well understood, if not of Things (IoT) perspective, where a big role is played by
outright xed in stone. This paper challenges the status quo, (wireless) sensor and actuator networks (SNs/WSNs).
advocating a novel kind of Infrastructure as a Service, in The outlook points to an outright constellation of stan-
order to stimulate a discussion on whats next for Cloud dalone devices and SNs alike, all going to be connected
computing, especially where it meets the Internet of Things. over Internet. Such ecosystem of geographically distributed
A rst-hand experience on any mass-market mobile gad- resources cant be left untapped, and all those resources
get, not to mention ordinary desktops, can readily provide should be made available for discovery and selection, ac-
clues about how ubiquitous Cloud-powered services already cording to provided functionalities, eventually cooperating
are. Still cheaper hardware and economies of scale dont autonomously towards a common goal.
play any role other than getting ever fatter clients in the Sensing and actuation resources in particular should be
pockets of wider pools of customers. part of the Cloud, not just as mere data endpoints, as
Even just talking about customers may help to reveal the envisioned by current mobile Cloud trends, but abstracted,
biggest shortcoming of the current wisdom about Clouds: virtualized and (administratively) grouped in Clouds, i.e. to
the supply chain stops right before the end-user. If we try to be dealt with along the same lines as computing and storage
think about it, this isnt the mainstream view when talking resources in typical IaaS fashion.

978-0-7695-5094-7/13 $31.00 2013 IEEE 60


DOI 10.1109/3PGCIC.2013.16
B. IoT D. Linking the IoT to Clouds
There are several ideas and meanings attributed to the Keeping in mind the foregoing description, a remarkable
emerging IoT concept. According to CERP [5], IoT can be intersection between SAaaS Clouds, volunteer contributions
dened as a dynamic, scalable, global network infrastruc- and mashups may just lie in the IoT, more specically in
ture, endowed with self-adaptive facilities based on inter- exposing underlying physical items through further abstrac-
operable communication protocols, standards, taxonomies tions according to thing-like semantics. The infrastructure
(folksonomies) and ontologies, where things, either physi- we are going to outline could indeed be the platform on top
cal or virtual, have in turn physical attributes, identities, and of which such abstractions would be implemented, where
virtual personalities, and employ smart interfaces, seam- thing handlers, pointing to physical items (e.g. documents,
lessly joining the information highway. cars, products, parts, etc.), are available for discovery, selec-
Things are thus poised to acquire active roles in social, tion and allocation, enabling the Things as a Service (TaaS)
information and business processes, enabled to interact with provisioning model. Our purview is reframed accordingly,
the external environment and among themselves by exchang- in terms of a Cloud of Things (CoT) [10] that, compared to
ing sensed data, and to react autonomously to physical world IoT and WoT (Web of Things) paradigms, bears signicantly
events, as well as to affect the environment they are plunged higher expectations than what mere interconnection and
into by means of triggers, and actuators acting upon, in the hyperlinking of things may yield.
end disclosing services, even without human intervention. Indeed the missing link could be a hypervisor of handles
C. Keys to the Cloud on things to manage the lifecycle of several non-exclusive
perspectives on the same (physical) item. On one hand,
When talking about Clouds its crucial to identify, and
theres a need to decline a handle into a certain context
focus on, the mainstream categories, as well as the layers at
according to the task under request, thus binding a subset of
which implementations get developed and operations occur
the pointed resources. On the other, a requirement is to book
in order to offer services. In particular, apart and above the
just enough capabilities, leaving the underlying devices
user- and developer-facing services, respectively Software as
open to concurrent exploitation, without further perspectives
a Service (SaaS) and Platform as as Service (PaaS), the real
overstepping on the scheduled ones, in pure IaaS fashion,
engine behind Cloud enablement is in Infrastructure as a
i.e. exposing elastic infrastructure from limited hardware
Service (IaaS) innovations and technologies. Hybridization
resources.
(e.g. mashups) and exploitation of SaaS and PaaS brings new
To implement this challenging vision, mechanisms and
services to wildly heterogeneous populations of customers,
tools for SAaaS, volunteer framework, resource mashup and
yet IaaS is a relatively recent development in computing, and
TaaS are needed. A tentative roadmap can be drafted iden-
a very powerful vector for new possibilities up the stacks.
tifying the following research and development milestones:
Clouds therefore become endowed with computing, stor-
age and sensing too, thus creating a pervasive infrastructure 1) SAaaS enablement: WSNs, smartphones or other de-
that is able to interact with the surrounding environment. vices, endowed with sensors/actuators, to be managed
Novel possibilities for contextualization and geo-awareness in a Cloud environment;
get unlocked by leveraging previously untapped infrastruc- 2) Volunteer involvement: application of volunteer com-
ture, i.e. sensors and actuators. Per naming conventions, as puting approaches to SAaaS node enrolment;
applied to virtualized computing resources, i.e. IaaS, and 3) Resource mashup: meeting special requirements with
storage resources, in terms of DaaS, short for Data as a the help of IaaS/DaaS Clouds;
Service, our approach could be dened with an acronym 4) Things abstraction and virtualization: disclosing the
such as SAaaS, i.e. Sensing and Actuation as a Service [11]. CoT by mashups according to the TaaS paradigm.
By nature and provenance of sensing resources, possibly The aforementioned phases would follow a chronologi-
involving mobile devices in volatile settings, the scenario cal/logical order but are not interlocked, i.e. they could par-
is prone to be highly dynamic, well beyond the typical tially overlap. In accordance with the roadmap, a blueprint
(i.e. xed) infrastructure exploited for Cloud computing and of the architecture implementing the functionalities above
storage services. Thus, a workable plan to address such identied is shown in Fig. 1, where the stack as a whole
issues suitably is to resort to the volunteer contribution is depicted, outlining the modules for a TaaS-based CoT.
model as an underlying approach. There is an IntraNode layer at the bottom, i.e. an execu-
So-called heterogeneous resource mashup, i.e. orchestrat- tion environment at the level of a single node, where to
ing resources of volunteer-based sensing-powered Clouds abstract away either embedded sensors, e.g. those available
together with those from mainstream computing and storage on a smartphone, or standalone ones, usually belonging
Clouds, whether for internal use or as public offerings, is to a network of uniformly capable devices (WSNs). At
among the prominent issues to be addressed in building such the IntraCloud/InterNode level lies every interaction among
a wide-ranging Cloud (eco)system. nodes belonging to a single (SAaaS) Cloud, to be generated

61
SaaS - IoT SAaaS module

CoT Services Apps BPaaS


Abstraction & Virtualization
Unit
Node Manager
P InterCloud - PaaS
h
a
s TaaS Provider
e Adapter
4 TaaS module
Device
P
h
a
s
e
IntraCloud/ Figure 2. SAaaS module
3 SAaaS Providers
InterNode
P VolunteerCloud
h Existing IaaS Existing DaaS Management
a
s Providers Providers
e
Autonomous Autonomous
In Fig. 2 an high-level view of the internal architecture of
2
module module the SAaaS module is depicted, subdivided into three main
P
h IntraNode SAaaS
module
SAaaS
module
modules:
a
s
e Devices
Sensor Network
Gateways i. the Adapter interfaces directly with sensing/actuation
1
devices and keeps track of connectivity, translating
commands and relaying them to the underlying physical
Figure 1. Architectural schema and modules endpoints, by means of device-native communication
protocols.
ii. the Node Manager is in charge of policy enforcement,
and exported to the upper layer, InterCloud, where interop- exclusively at node level but enhanced with autonomic
erability and federation among Clouds are the core issues features, e.g. self-optimization with regards to energy
a TaaS provider needs to deal with. More specically the efciency and self-conguration to enable sensor virtu-
highest layer should mash up heterogeneous (IaaS/DaaS) alization.
Clouds with SAaaS ones, interacting with the former through iii. the Abstraction and Virtualization Unit exposes the
industry-wide interfaces and standards. There are four core functional aspects of the node as services, relays sen-
modules in this architecture, from the bottom up: SAaaS, sor events to upper layers, and virtualizes underly-
Autonomous, VolunteerCloud Management and TaaS. ing resources, e.g. repurposing or aggregating sen-
Aim of this paper is to provide an overview of the CoT, sors/actuators to shape the functionality the node ex-
and related enabling technologies. Although we have already ports.
implemented solutions for a few problems, here we focus on
The SAaaS module therefore provides sets of sen-
the big picture and on the main inspiration about the CoT,
sors/actuators in the widest possible meaning (e.g. printers,
thus only providing pointers to previous efforts and ongoing
webcams, human interaction peripherals), or even arbitrarily
work we are carrying on, starting also from, and referring
complex ones, e.g. weather sensors as virtual resources
to, existing solutions and background.
that consist of basic temperature, pressure and humidity
To this purpose, following the roadmap heretofore iden-
sensors, i.e. originating from their composition. Moreover
tied, the architecture will be detailed below.
it supports hypervisor-like features, i.e. splitting one sensor
II. SA AA S ENABLEMENT into a multitude by tuning e.g. sampling parameters, or even
repurposing it, e.g. providing motion detection services with
At the bottom of the stack, the SAaaS module is needed a camera, in order to cope with the lack of positioning
for the abstraction of sensing and actuation resources, en- sensors.
hanced with policies and data handling. It exposes function- Starting from existing languages and models providing
alities to be provided at the level of what would represent formal descriptions of sensor systems and processes as-
a single node, i.e. a management domain, comprising either sociated with sensor observation (SensorML, Observations
a WSN [12], [4], [3], hidden behind a gateway, or a set and Measurements Schema within the OGC Sensor Web
of sensors embedded in a standalone device. A node, when Enablement [20]), the lower level of the above architecture,
talking about SNs, may turn out to be less easily identiable i.e. the adapter and the abstraction modules, has been already
when compared with the one-to-one mapping that naturally implemented into the CLEVER stack [14].
springs to mind when talking about smartphones and other The Node Manager and the Abstraction & Virtualiza-
personal devices: typically an SN may be made up of tion Unit are work in progress, also based on the W3C
thousands of sensors, yet exposing a limited subset of sinks, Semantic Sensor Network Incubator Group initiative, ex-
amongst other considerations the only components where tending SensorML-enabled syntactic level interoperability
nodal modules of the stack could be deployed.

62
Volunteer Framework
order to enforce self-adaptively the policies coming from
the VolunteerCloud Management module.

VolunteerCloud
Autonomous Management
APIs
Looking at the framework as a whole, it is also reason-

Module
able to leave room for fully autonomous volunteer-SAaaS
Reward QoS SLA Clouds, set with well-dened default policies (e.g. best-
System Manager Manager
effort), where the VolunteerCloud Management module is

Module
not strictly needed, thus leaving all the duties to be self-
managed by the Autonomous module.
SAaaS module The Autonomous module makes up a decoupled, two-
level, hierarchical autonomic management system in com-
bination with the SAaaS Node Manager, the whole being
Figure 3. Volunteer Framework
entirely deployed and running on the node. The former
works at node level, e.g. within a SN domain, whereas the
to a semantic level, through the development of both an latter tracks and enforces higher-level Cloud targets.
ontology for describing sensors and their data, and an The Volunteer Framework thus builds a volunteer-based
annotation framework for adding semantic metadata to the sensing Cloud upon the SAaaS module, and exports services
SWE standards. for mashups and TaaS, by providing the following compo-
However those standards do not yet provide a full-edged nents, spanning the two aforementioned layers:
semantic description of the sensor and the phenomena it i. Reward System: increases reliability and availability of
measures, a central theme in Section V. volunteer-borne services based on sensors/actuators, and
III. VOLUNTEER INVOLVEMENT keeps track of resource usage for management and
billing purposes;
The Volunteer framework is needed to consolidate ii. QoS Manager: implements the monitoring framework
volatile, ad-hoc resources and services, such as volunteer- for quality of service, at the level of a single Cloud;
borne sensors, in a Cloud setting. The core duties are iii. SLA Manager: enforces the SLAs ratied by the coop-
about techniques for mitigation of the effects brought about erating layers of the whole stack, e.g. TaaS, mashups or
by resource churn, where performance is largely dynamic, just peer Clouds.
lifespan is short, nodes are heterogeneous and mobile, and
information on their status is partial and often already out- In summary, the Autonomous module logically translates
of-date when the need arises. It follows that, while operating and tie-breaks reward, QoS and SLA autonomic policies
on largely unpredictable and unreliable resources, this layer on the SAaaS nodes, applying directives coming from the
provides increasingly dependable services to the upper ones. VolunteerCloud Management module. In order to achieve the
Decentralized, on-the-y selection of services requires best node conguration while satisfying most QoS, SLA and
dynamic binding, and involves active service tracking, mon- Reward-Credit system constraints, such policies have to be
itoring and rich metadata, including QoS/SLA metrics. locally coordinated.
Distributed broker agents coordinate to keep track of the In relation to these issues we are going to leverage the
energy prole, quality, availability, location and context of experience acquired, and work done, in the Cloud@Home
services, and pick resources according to criteria and con- project [7], [6], [8], aimed at implementing fully volunteer-
straints. Another signicant topic is about how information contributed Clouds that can interoperate with, and federate
and metadata about services (e.g. QoS, SLA) should be to, commercial Clouds, mixing aspects and ideas coming
handled and indexed, considering distribution, heterogeneity from volunteer approaches (credit systems, BOINC [1]) and
and likelihood to quickly get outdated, due to the steady P2P mechanisms (i.e. Jingle, as per XMPP enhancement
turnaround of volunteers. proposals) with the Cloud provisioning model.
Moreover, in order to increase the reliability and avail- With regards to the CoT, a critical task, up to the Au-
ability of services, a multi-dimensional reward system is tonomous module, is to provide support for identication
needed, motivating contributors to provide more predictable of things. The collaboration of node owners and adminis-
metadata and stay involved for longer. trators, providing specic input to augment knowledge on
The Volunteer Framework is logically structured into things with metadata, through semantic Web technologies
two modules, as depicted in Fig. 3: the VolunteerCloud [15], is needed in order to further the CoT agenda. This
Management and the Autonomous ones. The former stores requires a common and well-known ontology framework
and executes management strategies at the Cloud level by as a prerequisite, shared and agreed upon by both TaaS
means of a steady interaction with every node belonging owners/administrators and providers.
to the newly-formed (sensing) Cloud, while the latter is Still, higher-level mechanisms and tools have to be em-
to be deployed into each node of the SAaaS Cloud in ployed provider-side, e.g. implementing feature extraction

63
trustworthiness, billing and reward-credit systems, and non-
TaaS module functional properties as well, negotiated through tailor-made
Frontend SLAs.
A further high-level abstraction is needed in the TaaS
Things Abstraction
& Virtualization module, where mashed up resources are handled as things
in a CoT perspective, chiey requiring semantic efforts in
High-Level Manager
the assignment of thing-like meaning to resources. These
Cloud Driver
tasks belong to the Things Abstraction & Virtualization, as
described in detail in Section V.
At last we have the Frontend, providing tools for cus-
Figure 4. TaaS module architecture
tomer interactions with TaaS Clouds, collecting and relaying
requests issued by users, e.g. through a Web portal, imple-
menting access facilities for the Cloud, in particular related
services, complementary to user-generated metadata, in or- to identication, authentication and authorization, e.g. by
der to properly identify and describe things, as discussed in means of an infrastructure based on PKI and related tech.
Section V. A good starting point for resource mashup could be the
one by existing infrastructure providers, in terms of both
IV. R ESOURCE MASHUP IaaS (EC2 and the like) and DaaS (S3 or similar) therefore
The TaaS module deals with issues related to the upper exploring issues in mashing up heterogeneous resources. In
management duties for the overall infrastructure, and builds particular there are some interesting third-party solutions
on top of already formed (sensing/actuation, computing (Rapidscale, EMC2) implementing brokering, management
and storage) Clouds, enabling heterogeneous infrastructure and coordination activities.
mashups, in particular geared towards the CoT. On the interoperability and federation front, our re-
Location awareness, and other contextual metadata, are search unit is developing CLEVER, an innovative IaaS
poised to help deeply in the optimisation of computing, stack offering an advanced storage, messaging and presence
storage and sensing interCloud management, hence those framework, hinged on the XMPP family of standards, for
should be considered as potential key differentiators of the interconnection among administrative domains.
proposed approach for an interCloud solution that needs to V. L AST MILE TO THE C OT:
work in a distributed fashion (e.g. policy-based), so as to ABSTRACTION AND VIRTUALIZATION
provide support for heterogeneous Clouds.
Issues to be addressed here are: interoperability, abstrac- The main theme of this efforts is in relation to the Internet
tion, virtualization and management of mashed up resources of Things, and lies in what we identied as the Cloud of
for TaaS, SLA, QoS, security (e.g. trustiness, privacy, iden- Things, or CoT, paradigm.
tity management), billing and accounting, many of which Indeed the underlying infrastructure, as heretofore de-
are to be tackled across the whole framework. scribed, is geared towards building a CoT framework, since
In Fig. 4 a draft outline of the structure of the TaaS things, per the IoT denition, RFID-tagged or otherwise
module is depicted, composed of four main parts: the Cloud featuring smart sensors/actuators, are typically mobile, and
Driver, the High-Level Manager, the Things Abstraction & a stack of this kind is poised to provide all the enabling
Virtualization and the Frontend. technologies to tackle this vision.
The Cloud Driver provides interfaces to volunteer Clouds, It goes without saying that such a framework would
and non-volunteer ones as well, collects and manages infor- turn out to be very valuable even without a CoT level.
mation about Clouds availability and provided services, in Yet the latter can be a killer application for this kind of
terms of both functional and non-functional properties, e.g. infrastructure, where the pay-off is an IoT scenario, enriched
QoS, costs and reliability. by enhanced context and unlocked degrees of freedom for
Its main duty is to provide interoperability among vol- discovery and allocation of resources, setting the stage for
unteer and proprietary Clouds alike, and to cooperate with a marketplace of things.
the High-Level Manager to carry out heterogeneous resource As outlined in Section IV, the last step along the way
discovery. to the CoT goes through the abstraction of mashed up
resources, by virtue of renements, specications and im-
The High-Level Manager in turn is a key building block of
plementations based on the notion of thing, semantically
TaaS: it manages incoming requests of mashed-up cross-
tagging the underlying items.
type resources by splitting them into simpler single-kind
requests, to be dispatched to the corresponding SAaaS, IaaS In the TaaS module the Things Abstraction & Virtualiza-
or DaaS providers, thus can be compared to a Cloud broker. tion block has to enable search of things satisfying user-
Other issues to be considered are related to security, QoS, dened requirements, e.g. devices with certain monitoring

64
based on a set of OWL DL [2] ontologies, to be bootstrapped
Things Abstraction &
Virtualization
and populated [9] through semantic partitioning [21], taxon-
omy mining and instance extraction.
Management
The ontologies should be coupled with a query metasys-
Virtualization tem [17], [22], hinged on frame-based languages [16] with
Manchester syntax, in order to avoid caring about domain-
Mapping & Transformation
specic syntax quirks. While crucial, domain ontologies
Detection & Identication are only part of the answer here, as even narrowing the
scope for identication to a limited selection of predened
Mashed-up Resources categories, i.e. according to mainstream use-cases, cannot
entirely eschew cross domain-spanning relations. In this
context, ontology alignment [13] would come to the rescue,
Figure 5. Things Abstraction & Virtualization module architecture possibly aided by upper (foundation) ontologies [19] as glue,
in the role of ancestors along the semantic hierarchy.
As an example at this stage, a few smartphone owners
features, vehicles inside specic areas, people (or things) could volunteer for being part of the Cloud, and the app UI
whose positions or trajectories follow a predened pattern, would ask info to the participant in an opt-out fashion. Yet
and so on, according to the application domain, per request external venue-based check-in machinery (i.e. at the stadium,
specication. concert hall, gas station) could still be triggered unaided as
Moreover pre-processing and ltering of data are among soon as a proximity event res, and tag phones, or more
its core duties, in order to trigger custom algorithms and accurately, the embedded sensors onboard, accordingly, by
provide new value added services, e.g. goods tracking, their identiers in the Cloud. In both situations, lightweight
trafc control, emergency management, remote monitoring. pattern recognition methods could be applied to the envi-
Therefore ontologies, and the taxonomies they encode, are ronment the mobiles are plunged into, in order to enrich
required to label things. acquired metadata.
An abstraction layer over the mashed up resources can be Up one level the Mapping & Transformation step takes
decomposed in a Detection & Identication module below, place, where the abstraction mechanisms come to fruition.
and a Mapping & Transformation one above, as depicted in Here things, or individuals according to semantic termi-
Figure 5. nology, make up both the inputs and outputs of this block,
Its useful to pinpoint use cases based on smart cities undergoing the following operations, to name a few:
scenarios, thus following through with examples, in order
mapping one thing to belong to more classes;
to lay out a detailed description of these layers. Lets focus
decomposing a thing into its parts;
on (urban) trafc control and major events management as
morphing the class of an individual in another one more
driving applications of the aforementioned scenario.
apt (e.g. higher popularity) to be matched to requests
From the bottom up, theres the Detection & Identication for service, albeit under well-dened constraints, for
module rst, tasked with detection of features that can help instance a sibling relationship.
identifying things, i.e. collecting both inner and (contextual)
user-provided information relayed from the lower levels, as Other typical use cases would include specialization or
well as autonomously generated relevant metadata by means generalization of the classes things belong to, tantamount
of deployed (or builtin) check-in mechanisms, or traps, to to semantic super-/sub-classing, by traversal of the ontology
signal the availability of new clues for otherwise already graph.
tagged things. Furthering our scenario-based blueprint, an identity map-
One more avenue is top-down feature extraction, central- ping could be established in order to relay the output of
ized yet scalable, i.e. hierarchical and inherently level-local, the classication (Identication) block transparently to the
services for pattern recognition out of historical and real- Virtualization level, in the simplest case. On the other hand,
time samples. Moreover, metadata collected through one of already stretching a bit the usefulness and complexity of
these channels could be validated against info from other the use case, an aggregation or composition of things could
sources. be the output. Pushing the envelope, recursive relationships
All these techniques should therefore be viewed as may help generating full-blown ecosystems. Summing it
matched parts of the whole machinery, as setting those up, whichever the transformations or the complexity, even
mechanisms in motion at the earliest, i.e. node enrolment at this step one or more reasoners [23] are needed, endowed
stage, is a precondition for proactive collection of relevant with relevant knowledgebases [18].
clues to be passed on to the Identication step, to be in In line with the examples heretofore developed, we can
turn carried out by a semantic inference engine, for instance think of situations like:

65
a commuter that becomes also a member of an audience to ease the scaling and timing issues, a probabilistic kind
(mapping); of matching [15] for discovery needs to be achieved, by
a group of people that gets tagged as a crowd, for combining approximate, or non-deterministic as a last resort,
instance a ash mob (aggregation); classication with thing (class) look-up supported by e.g.
a car unbundled into its parts, a family into its members, requirements substitution. At lower levels the prevalent
or a theater into its different areas (decomposition); volunteer nature of contributors would naturally lead towards
a trafc jam that just melts into an uneventful ow P2P methods (e.g. distributed hash tables) for indexing and
(morphing); discovery duties. Yet such an approach would be a good t
an unidentied vehicle whose description gets further also at the TaaS layer, if we consider that, however (partly)
rened into a SUV (specialization), and so on. centralized, the indexing and labeling of the things would
Still treading the analogy between IaaS and TaaS, whereas happen across families of resources, administrative domains
the Abstraction level provides one or several representations or Clouds. Since TaaS allocation and business logic policies
for a single item or for a whole environment of things, en- have to be mainly mapped into lower level policies on
coded as semantic cardinality, the Virtualization layer could mashed up resources, in such case the Management system
indeed be thought as exposing functionalities equivalent to implements a kind of wrapper of incoming requests into
a hypervisor of handles on the abstracted things. In this mashed-up ones, from that point onwards managed by High
way it is able to arbitrate and manage instantiation and Level Manager of the TaaS module, as discussed in Section
lifecycle of multiple perspectives on the same physical IV.
item, i.e. sets of handles abstracted from (e.g. sensing) nodes Recapping, in a top-down fashion, the management layer
belonging to the semantically dened individual. This allows processes free-form requests probabilistically matching ap-
to contextualize the whole thing, i.e. the generated set of its plication domains, while the virtualization one below instan-
resources, to pending tasks, binding a subset of resources to tiates perspectives on things in a deterministic way, based
the request per the application domain. on matched domains as relayed from above.
As an example, a request for monitoring the speed of vehi-
VI. R EADY FOR C OT?
cles on a certain road, along with their spatial conguration,
for instance for trafc jam forecasting and prevention, would Theres no doubt that Cloud computing and IoT are
translate into instantiation of perspectives on the enclosed among the trendiest topics in ICT. Here we tried to address
vehicles. Each handle therefore represents a perspective unexplored ideas at their intersection, where things may
on a car (or a group of), in this case just made up of be provided dynamically according to a service oriented
geopositioning and motion sensors allocated from onboard paradigm, thus shifting the boundaries towards a Cloud of
devices (e.g. in-dash) as well as volatile ones, such as those Things, i.e. employing the Cloud provisioning model to meet
on the mobiles of passengers. user needs according to guaranteed service levels.
Bringing home the point, virtualization means, by IaaS The CoT thus can only be built atop a platform suited for
denition, exposing more infrastructure than the available the management of:
hardware, in order to keep the pipe full at all times, i. sensing (and actuation) resources exposed as a ser-
concurrently satisfying as many requests as possible, thus vice (SAaaS) from mobiles and SNs alike, addressing
squeezing maximum value from the setup expenses, while volatile availability and reliability, especially when mo-
keeping the underlying constraints unbeknownst to the cus- biles are involved, by means of volunteer computing-
tomer. inspired techniques;
In the CoT setting, virtualizing means guaranteeing non- ii. mashed up resources, provided through heterogeneous
exclusive access to the devices exposed in a perspective (IaaS, DaaS and SAaaS) Clouds, also addressing issues
where possible, as long as the Abstraction layer exports related to interoperability and federation;
resources satisfying the minimum requirements under the iii. things, by means of suitable ontologies and semantic
constraints emanating from the request for service, hence approaches, to be adopted en masse and standardized
justifying once more the need for transformations. upon by most users, customers and providers, in order
Keeping with our example, on the aforementioned sensors to label, map and transform mashups in a TaaS fashion.
sampling would be tuned in order to export enough func- In this way TaaS unlocks pervasive, innovative and value-
tionality for the task at hand, leaving the same underlying added applications under the guise of CoT to any customer,
devices open to further exploitation through request multi- at the same time a potential provider as well, thus disclosing
plexing for new perspectives, made up of handles compatible an open marketplace for things. These new avenues could
with the allocations already in place. require input from, or be input for, innovative research on
At the top of the whole module a Management system revenue models and business methods, behavioral and social
would deal with discovery, allocation and business logic aspects (i.e. psychology, sociology, public relations), legal
mechanisms, as per best practices for plain IaaS. In order issues (service level agreements, licensing, privacy), service

66
engineering (human-machine interaction, visualization, de- [12] Salvatore Distefano. Evaluating reliability of wsn with
sign), and other elds as well. sleep/wake-up interfering nodes. Int. J. Systems Science,
44(10):17931806, 2013.
R EFERENCES
[13] Marc Ehrig. Ontology Alignment: Bridging the Semantic Gap,
[1] David P. Anderson and Gilles Fedak. The computational and volume 4 of Semantic Web And Beyond Computing for Human
storage potential of volunteer computing. In CCGRID 06, Experience. Springer, 2007.
pages 7380. IEEE Computer Society, 2006.
[14] M. Fazio, M. Villari, and A. Puliato. Sensing technologies
for homeland security in cloud environments. In Sensing
[2] Sean Bechhofer, Frank van Harmelen, Jim Hendler, Ian Hor-
Technology (ICST), 2011 Fifth International Conference on,
rocks, Deborah L. McGuinness, Peter F. Patel-Schneider, and
pages 165 170, 28 2011-dec. 1 2011.
Lynn Andrea Stein. OWL Web Ontology Language Reference.
W3C, dean, mike and schreiber, guus edition, February 2004.
[15] Sara Hachem, Thiago Teixeira, and Valerie Issarny. Ontolo-
gies for the internet of things. In Proceedings of the 8th
[3] D. Bruneo, A. Puliato, and M. Scarpa. Dependability Middleware Doctoral Symposium, MDS 11, pages 3:13:6,
analysis of wireless sensor networks with active-sleep cycles New York, NY, USA, 2011. ACM.
and redundant nodes. In Proceedings of the First Workshop on
DYnamic Aspects in DEpendability Models for Fault-Tolerant [16] Matthew Horridge, Holger Knublauch, Alan Rector, Robert
Systems, DYADEM-FTS 10, pages 2530, New York, NY, Stevens, and Chris Wroe. A Practical Guide To Building
USA, 2010. ACM. OWL Ontologies Using The Protege-OWL Plugin and CO-
ODE Tools Edition 1.0, August 2004.
[4] Dario Bruneo, Salvatore Distefano, Francesco Longo, Anto-
nio Puliato, and Marco Scarpa. Evaluating wireless sensor [17] Dimitrios A. Koutsomitropoulos, Ricardo Borillo Domenech,
node longevity through markovian techniques. Computer and Georgia D. Solomou. A structured semantic query inter-
Networks, 56(2):521532, 2012. face for reasoning-based search and retrieval. In Proceedings
of the 8th extended semantic web conference on The semantic
[5] Cluster of European Research Projects on the Internet of web: research and applications - Volume Part I, ESWC11,
Things (CERP-IoT). Internet of Things Strategic Re- pages 1731. Springer-Verlag, 2011.
search Roadmap. Technical report, European Commis-
sion - Information Society and Media DG, http://www.grifs- [18] Holger Neuhaus and Michael Compton. The Semantic Sensor
project.eu/data/File/CERP-IoTAccessed May 2013 2009. Network Ontology: A Generic Language to Describe Sensor
Assets. In AGILE Workshop Challenges in Geospatial Data
[6] V. D. Cunsolo, S. Distefano, A. Puliato, and M. Scarpa. Harmonisation, 2009.
Cloud@home: Bridging the gap between volunteer and cloud
computing. In 5th International Conference on Intelligent [19] Ian Niles and Adam Pease. Towards a standard upper
Computing, ICIC 2009, LNCS, pages 423432. Springer, ontology. In Proceedings of the international conference on
2009. Formal Ontology in Information Systems - Volume 2001, FOIS
01, pages 29, New York, NY, USA, 2001. ACM.
[7] V.D. Cunsolo, S. Distefano, A. Puliato, and M. Scarpa.
Volunteer computing and desktop cloud: The cloud@home [20] C. Reed, M. Botts, J. Davidson, and G. Percivall. Ogc(r)
paradigm. In Network Computing and Applications, 2009. sensor web enablement: overview and high level achhitecture.
NCA 2009. Eighth IEEE International Symposium on, pages In Autotestcon, 2007 IEEE, pages 372 380, sept. 2007.
134 139, july 2009.
[21] Julian Seidenberg and Alan Rector. Web ontology segmen-
[8] A. Cuomo, G. Di Modica, S. Distefano, A. Puliato, M. Rak, tation: analysis, classication and use. In Proceedings of the
O. Tomarchio, S. Venticinque, and U. Villano. An sla-based 15th international conference on World Wide Web, WWW
broker for cloud infrastructures. J. Grid Computig, 11(1):1 06, pages 1322, New York, NY, USA, 2006. ACM.
25, 2013.
[22] Smith M. Sirin E., Bulka B. Terp: Syntax for owl-friendly
[9] H. Davalcu, S. Vadrevu, S. Nagarajan, and I.V. Ramakrishnan. sparql queries. In 7th OWL Experiences and Directions
Ontominer: bootstrapping and populating ontologies from Workshop, San Francisco, 2010.
domain-specic web sites. Intelligent Systems, IEEE, 18(5):24
33, sep/oct 2003. [23] Peter Yeh, Bruce Porter, and Ken Barker. Mining trans-
formation rules for semantic matching. In ECML/PKDD
2nd International Workshop on Mining Graphs, Trees, and
[10] S. Distefano, G. Merlino, and A. Puliato. Enabling the
Sequences, pages 8394, 2004.
cloud of things. In Sixth IEEE International Conference
on Innovative Mobile and Internet Services in Ubiquitous
Computing, IMIS, pages 858863. IEEE, 2012.

[11] S. Distefano, G. Merlino, and A. Puliato. Sensing and


actuation as a service: A new development for clouds. In
11th IEEE International Symposium on Network Computing
and Applications, NCA, pages 272275. IEEE, 2012.

67

You might also like