A Testbed for Experimenting IoTbased Big Data Applications

Outline
1. Introduction
2. Use case (Traffic Domain)
3. Challenges
4. Aim and objectives of our research (Motivation)
5. Contribution of the paper
6. Related work (Research papers)
6.1 Design and Deployment of an IoT Application Oriented Testbed
6.2 loT Lab: towards co-design and loT solution testing using the
crowd
6.3 A survey on facilities for experimental internet of things research
6.4 Smart Campus: A user-centric testbed for Internet of Things
experimentation
7. Proposed Testbed
8. Generic Architecture Diagram
9. Architecture Implementation
10. Evaluation and Results
11. Conclusion and Future Work
1. Introduction
// seeking your guidance
2. Challenges
// seeking your guidance
3. Aim and the objective of our research
// seeking your guidance
4. Contribution of the paper
Implementing the idea of an end to end system consisting IoT and big data
technologies is a challenging task due to the integration of different multiple
communication protocols, connectivity standards, frameworks and platforms.
It creates interoperability issues among different technologies. Hence,
deployment of such a large scale end to end system becomes very time
consuming and difficult due to the high learning curves, involvement of
multiple technologies on different layers, etc. Thus, the idea of developing a
testbed is helpful to realize the feasibility of a proposed end to end system.
This paper proposes a testbed architecture for experimenting IoT based big
data analytics. In order to realize the actuality of a testbed we have
implemented two use cases. [1]. Traffic scenario and [2]. Smart lighting. The
first use case is about traffic scenario, where using the multiple emerging
technologies visualization has been generated which converts raw data into
useful knowledge. In the second use case, real motion detection sensors are
deployed to enable the smart lighting in the building. In this use case, the
system is analyzing the energy consumption of the lighting set up in the
building.

5. Related Work (Research papers referred)

5.1: Design and Deployment of an IoT Application Oriented Testbed

The paper represents the idea of WoTT (Web Things testbed) for the testing
and Designing real world innovative IoT products and application that solves
the problem in different areas such as healthcare, transportation, security,
etc. The global reach and extreme heterogeneity of the Internet of Things
present major application development challenges. Here, Testbed provides a
stable, open, dynamic, and secure infrastructure to simplify application design
and testing. This paper presents two use cases of actually deployed testbeds
in the field of energy consumption and environmental data collection.

Example-1: IoT-Lab. It provides a very large scale infrastructure with more

than 2700 wireless sensor nodes spread across six different cities in France
and is used to test protocols at the link and network layers and to collect
performance results (Energy consumption or packet delivery ratio).
Example-2: SmartSantander- There are 20,000 sensing nodes deployed
across Europe for sensing parameters (temperature, CO, noise, light, car
presence, etc.) (Environmental data collection).

What is IoT hub? How is it useful? The IoT Hub is not just required for interaction
and interoperability, but its presence extends the IoT network and increases
its capabilities by simplifying and hiding complex and important tasks such as
service discovery and routing. As per the authors WoTT (Web of things
testbed)s main goals are to [1]hide low-level implementation details,
[2]enhance network self-configuration by minimizing human intervention,
[3]transparently and simultaneously manage multiple protocols and
platforms, and [4] provide a platform for the design and testing of human
object interaction patterns

IoT hub - WoTT does not simply enable communications between IoT actors; it
constitutes a uniform super-entity able to provide enhanced functionalities
that go beyond the mere union of its components features. To achieve this
super-entity status, WoTT uses the various communication technologies in the
IoT Hub to bridge and merge together several networks into a single IP
network

Challenges in Developing WoTT

The main challenges encountered in WoTT deployment have been design
related: how to define different elements and their functionalities, represent
different resources and their relationships through suitable hypermedia, and
maintain compatibility with standards

Benefits of WoTT
The efforts devoted to WoTT design, the IoT Hub, and the use of standards
have simplified the deployment process, making the integration of all different
elements straightforward, notwithstanding their heterogeneity.

5.2: loT Lab: towards co-design and loT solution testing using the
crowd
This paper talks about an IoT Lab project which is about researching the
potential of crowdsourcing as an extension to the traditional IoT
infrastructure. This project proposes the novel approach for implementation
and testing of solutions with the involvement of crowd in process. Here, user
can participate in the experiment through contributing with sensory data and
knowledge. The goal is to engage the crows in achieving goals of innovative
products, process optimization and problem solving in different areas and
industries such as design, science and health, IT, finance, entrepreneurship,
social enterprise and many more. In this approach, the crowd generally
contributes with knowledge, time, feedback, ideas, resources, or even funding
to a project.
 Figure-1 Architecture for IoT lab platform

In this crowd driven approach, the crowd proposes a potential experiment for
IoT lab platform. Users work as sensing nodes and provide information
through the smartphone application. These ideas and information are
collection and the crowd assess this information or ideas. The selected idea is
implemented by the researchers and again evaluated by the crowd. This
iterative process continues and generates more impactful research and
results.

The paper also proposes unique model of the Testbed as a Service (TBaaS).
The Testbed as service model grants all integrated testbeds the ability to be
accessible in a uniform and transparent way by using cloud technologies. All
the necessary APls and underlying mechanisms that meet the requirements
and the architecture design of loT Lab platform will make all testbeds function
as a single loT meta-testbed providing a TBaaS cloud layer. In the TBaas, The
federated virtualized resources will be accessible via a range of technologies
that are supported by the vast majority of modern devices and systems. It will
also provide new business models for better development of Testbed
experiments.

5.3: A survey on facilities for experimental internet of things

research
The idea of IoT world has grown to multiple dimensions enclosing sensor
networks which can provide solutions and goal oriented intelligence to the
widespread things via network or internet. In spite of different advancement in
technology, challenges related to evaluation of IoT solutions under real
scenarios in real world experimental deployments still hinder their maturation
and significant rollouts. IoT is multidisciplinary domain that covers large
number of challenges and areas from purely technical (routing protocols,
semantic queries,) to a mix of technical and societal (security, privacy,
usability), as well as business and social areas. Existing and potential IoT
application are equally diverse. Designing a system that can efficiently and
effectively support such a large range of applications and be compliant with a
plethora of often contradicting requirements as well as integrating all required
components and technologies is a complex task. Simulations, as an important
phase during this process are useful for developing further understanding of a
system. However, they suffer from several imperfections [1] as they make
artificial assumptions on radio propagation, traffic, failure patterns, and
topologies. What makes it particularly difficult is the strong dependency of IoT
systems on real-world processes that are often result of complex systems-of-
systems interactions and extremely difficult to model accurately. To design
robust applications, developers need appropriate tools and methods for
testing and managing their applications on real hardware in large-scale
deployments. In the early days of IoT research, availability of smart devices
was limited. Although recent advancement have increased their availability at
lower costs. Although experiments were mainly small-scale and conducted in
research laboratories, they allowed for an improvement in understanding the
impact and limitations of real hardware on performance of protocols and
design choices. However, the daunting logistical challenge of experimenting
with thousands of small battery-powered nodes is the key factor that has
greatly limited the development of this field.

Overcoming the technical challenges and socio-economic barriers of a wide-

scale IoT deployment in our daily lives requires a thorough, practical
evaluation of IoT solutions using interdisciplinary, multi-technology, large-
scale, and realistic testbeds. The ability to test IoT solutions on a larger scale
and outside of research labs, i.e., in real environments and with real end users
has been considered only recently. These new testbeds aim to design and
deploy experimentation environments that will allow for:
[1] The technical evaluation of IoT solutions under realistic conditions,
[2] The assessment of the social acceptance of new IoT solutions, and
[3] The quantification of service usability and performance with end users in
the loop.

Testbed scopes:
 Figure-2: Testbed scopes

When building an IoT testbed, one key consideration is its scope. In terms of
the technological scope, a testbed can be single-domain if it addresses only a
particular type of IoT technology, e.g. wireless sensor network or RFID
devices. A multi-domain testbed is one that combines different IoT
technologies into a common experimental facility. The latter one is
increasingly important in addressing the IoT heterogeneity requirements.
Testbeds can be realized either indoor (Motelab) or outdoor (CitySense, Oulu
smart city, SmartSantander), thereby influencing the choice of the underlying
hardware architecture and software mechanisms. Indoor installations often
provide easier access to power and cabling for the testbed control and
management, making them easier to manage and more available. In contrast,
outdoor testbeds often rely on wireless connections, requiring additional
mechanisms in order to increase reliability. They also have to be protected
against malicious physical access and other threats that do not occur as
frequently in controlled environments. Furthermore, testbeds can be realized
either as permanent installations that are constantly available or portable
ones, usually temporarily deployed for measurement campaigns.
Testbeds can differ based on the application domain they are designed to
investigate. Generic testbeds such as Motelab, WISBED or SenseLab enable
experimentation research of communication protocols at different levels of the
protocol stack or application level algorithms that are mostly independent of a
particular application or service. More recently, domain-specific testbeds have
emerged that allow the evaluation of services and applications in a real
environment and from the end-user perspective.
5.4: Smart Campus: A user-centric testbed for Internet of Things
experimentation
Abstract Using IoT one can observe longitudinal (for a long time) observation
of Human behavior and Understand behavioral patterns that can inform
further IoT technology design. Currently, Experimentation with IoT
technologies is predominantly carried out in lab-based testbeds.

Problem - There is an emerging need for increased realism of the

experimentation environment, as well as involvement of real end users into
the experimentation Lifecycle.
What is SmartCampus? - a user-centric experimental research facility for loT
technologies. Compared to Lab based Testbed, SmartCampus deeply embeds
heterogeneous loT devices as a programmable experimentation substrate in a
real-life office environment and makes flexible experimentation with real end
users possible.

Problem exposure/Introduction Among various desired properties discussed

in the paper realism of experimentation environment, loT device
heterogeneity and real end user involvement is an important dimension to
improve experimentation life cycle upon existing lab based IoT testbeds.
-Realism connects experimentation conditions close to operating conditions
where final solutions are expected to deploy and helps to detect design flaws
or imperfections earlier and evened out, results in reducing the cost of roll out
and maturation time.

Figure: 3 Architecture of Use centric Test bed for IoT

Experimentation
Results Created an activity heat map show the no. of active devices at a
particular time for a data of a week. They had generated Social graph and
identified user/Device highest centrality during the day.

Implementation - Deployed in a real world office setting which spans an entire

building, SmartCampus overcomes the shortcomings of existing lab based
testbeds in the above-mentioned dimensions, while offering the same level of
control and configuration flexibility. They have IoT nodes installed on each
desk of work area sending data of noise, temperature, light and motion to
cloud server via Gateway, using collected data notify the user about energy
consumption on dashboard and allow the user to turn off device remotely via
an android app.

6. Proposed Testbed

7. Generic Architecture Diagram

 Figure-4 Generic Architecture for Big Data Analytics
Testbed

8. Implemented Architecture

Figure-5 Architecture implemented for Traffic Use case

In order to realize feasibility of the proposed testbed, traffic data simulator is

implemented with the help of our testbed, which is one of the use cases for
the proposed testbed. There is legitimate issue of authenticate traffic data for
the simulation and research purposes. Thus, analyzing and observing many
available traffic data sets, we have simulated traffic data for the
implementation of the desired use case. Simulated data (assumed as real
time traffic data) is being generated through node.js script. Simulated traffic
data is being published by MQTT publisher located on top of the raspberry pi
module. This published data is sent to the remotely located MQTT broker.
Now, An MQTT subscriber on another remotely located server subscribes to
the data located in the queue of MQTT broker. These real time simulated data
is sent through Kafka producer to Kafka Consumer. Kafka is feasible option for
real time streaming big data. Kafka consumer inject this data into the
database which further is used for visualization purposes. R platform uses this
database to request queries for visualization on dashboard. A package named
Shiny is used along with R for interactive dashboard.
9. Evaluation and results

Use Case: As mentioned earlier in the section, from the thorough research of
different traffic datasets we have generated simulated traffic readings for the
development of a traffic scenario use case. The simulated data from the
Raspberry pi (MQTT Publisher) is published to the MQTT subscriber through
the MQTT broker. Here, MQTT is used as it is lightweight protocol for M-to-M
(Machine to Machine communication). Especially in the case of sensors, it is of
utmost use as these embedded device environment operates under scarce
resource conditions. Not just this, but MQTT also considers bandwidth and CPU
limitations being the lightweight publish/subscribe messaging protocol. The
integration of Kafka is useful for the scalability purpose considering the huge
flow of streaming data. The Kafka is also another publish/subscribe messaging
system used for streaming big data. This real time streaming data thus stored
into the database from which traffic data values can be retrieved for the
visualization purposed with firing queries. Here, we are showing the
generated results with the help of Shiny package in R; which lets user interact
with the desired results.
In figure-4, an R dashboard window gives the interactive visualization for the
end user. Here, the simulated data consists of the type of vehicles present on
the road. An end user can get to know the total number of vehicles by
different filters such as date, location (A particular route) and hour range. A
user can observe the number of vehicles (with their types) passing by with the
filter of date and time.
In figure 5, user can compare vehicle flow on different routes filtered by date
and hour range. This real time streaming data is visualized to realize the
density of vehicular traffic on different routes on a given date and time.
In figure 6, an R window lets user know the vehicular density on different dates
(filtered by hours range) for the same location (Route). This is again achieved for the
real time streaming data. Here, end user can see the vehicular traffic condition on
different selected dates and for a specific hour range on the same location (route).

10. Conclusion and Future Work

A testbed is a useful solution for checking the feasibility of a concept or an idea. Not
just this, such testbeds helps the research fraternity to contribute to the real world
problems with the emerging and latest technologies at a rapid pace. The existence of
such open source projects helps in reducing significant efforts in the configuration
and integration of multiple technologies and frameworks. Undoubtedly, there are
issues regarding the authenticate datasets for research and analysis purposes.
However, there had been availability of accurate and authenticate datasets, the
proposed system can perform for well for the required analysis. Here, we have
implemented two use case: 1.Traffic scenario and 2. Smart lighting. This is a small
scale implementation with significant success from the whole system point of view.
This can be scaled further with more number of sensing nodes deployed on the
required locations.
---------------------------------------------------------------------------------
---------------------------Others things regarding testbed:
Test bed is an environment which is configured in accordance to meet the identified
test goal for the application/product .A test bed is a platform which allows:
1. Experimentation (Use of different technologies)
2. Evaluation (various use cases)
3. Testing of service usability (interactive visualization)
4. Testing of technical stability
Test beds are needed to understand a framework for experimentation of new
technologies under realistic operating conditions. It is a development environment
(or an environment set-up) which is useful for conducting rigorous and replicable
testing of solutions, concepts, theories and new technologies. Test beds are
developed to realize the actual scenarios with the freedom of wide experimental
possibilities (Contrary to actual scenarios which are more confined in nature)

Using A testbed one can:

1. Implement specific user scenarios (Traffic, Energy Saving, Healthcare, etc.)
2. Experiment with connecting different technologies together in a non-production
environment
3. Build final project requirements using test results
4. Discover opportunities for products and services outside the initial scope

In the context of IoT and Big Data Analytics

Any business (or research) looking for IoT implementation have some concerns
regarding cost, Efficiency, performance, etc. The idea of testbed helps in resolving
these concerns prior to the actual implementation. IoT based Big Data system
requires storage and processing infrastructure (Expensive), Sensors deployment,
maintenance, etc. Test bed helps in reducing the time and cost of making this
infrastructure and make the development much faster and easier. Due to the factors
such as interoperability and difference of latency and storage among different
technologies, it becomes very imperative to use the loop feedback control system.
Once the system is deployed, it may become complicated to modify the
implemented solutions. Therefore, a development of testbed is of paramount
importance for developing fairly large and complex system with integrating multiple
technologies from IoT, big data, cloud computing, etc.