The aim of this project is the virtualization of “things” to add
cloud based functionalities to a constrained IoT device. Building
smarter devices using an assembly of the existing ones is
achieved by message exchanges in a micro-cloud. A device
virtualization is featured by its input and output interfaces. The
proposed assembly strategy relies on putting interfaces together
and the definition of a richer component.
This effort focuses on three main topics:
• Makes connected objects highly available.
• Manage spatial distribution of composed connected objects.
• Handle and manage transactions on a collection of connected
LIMITATIONS & PROBLEMS
Mehdi Tazi firstname.lastname@example.org
Charif Mahmoudi email@example.com
Fabrice Mourlin firstname.lastname@example.org
Fabrice MOURLIN - 00 33 6 15 05 15 51
Université Paris-Est Créteil Val de Marne,
61 avenue du Général de Gaulle 94010 Créteil Cedex
The connected “things” needs to deliver high availability and
accessibility on top of a drastic resource limitations. Three main
limitations face the IoT systems:
• The battery autonomy: nowadays, the batteries autonomy
limits the active duration of the connected devices before any
human intervention. Their mobility is also affected as they
need to be connected to a charging sector or an electrical
• The processing power: the cost is the main driver of CPU
architecture embedded in the IoT devices. As they are usually
limited in processing power, many CPU-hungry functionalities
cannot be deployed on those devices.
• The storage capacity: IoT devices are not designed to store
data, their main propose is to collect data and send it on the
network. That prevents the possibility to deploy complex
filters on data that needs significant data caches.
• The network security: Usually poor in term of resources, the
IoT devices cannot host advanced security strategy. Their
exposition to the network makes this lake of security a real
• The data privacy: As the devices management can be limited
to device discovery, all the data are collected and stored in a
remote platform before the processing. Meanwhile, the privacy
of the data is not ensured especially on shared platforms.
Mehdi TAZI†‡ , Charif MAHMOUDI† , Fabrice MOURLIN†
IoT Virtualization in Micro-Clouds
The first goal focuses on by-passing the IoT mobile devices
hardware limitations by :
• Enhancing the connected objects lifetime : Inspired from the
mobile cloud computing (MCC) offloading techniques, the
batteries autonomy of the connected devices can lasts longer.
• Empowering the processing : Delegate and distribute the
processing from the IoT devices to micro-cloud nodes to
virtually add CPUs to the Devices.
• Increasing the storage capacity : by-pass the devices storage
limitations by offloading the data on distributed files systems.
• Identify a cloud stack offering the needed features
• Compute, storage and networking nodes
• Lightweight and reliable messaging system
• Multi-tenants and high available infrastructure
• Build and integrate Things Virtual Machine (TVM) in the stack
• Hypervisor for isolated VCT
• Design an IoT communication and composition protocol
• Composition protocol for building CVCT
• Communication protocol between CT, VCT and CVCT
• Benchmarking to validate the critical properties
• High availability, fault tolerance, resiliency, …
Paris-Est Créteil Val-de-Marne University, LACL Laboratory, France
‡ OCTO Technology
CT : Connected Thing
VCT : Virtual Connected Thing
CVCT : Composed Virtual Connected Thing
The second goal focuses on the delivery of the services provided
by the “things”. Still the integration within an IoT system is based
on ad-hoc components. The delivery of well structured IoT
services as a service offers a new paradigm for building IoT
systems. After the successful impact of this approach on regular
software, enabling “thing” as a service extends the Application
Capacity Type (ACT) of the cloud to the IoT world.
Extending the ACT to the IoT world brings many benefits :
• Capacity Improvement : using the micro-cloud scalability, the
“things” resolves complicated problems in short time for which
it wasn’t dedicated for at first such as image or voice
reorganization and databases management.
• Ease of maintenance, upgrade and portability : because of
the “VCT” nature, the “things” runs within isolated
environments, which permits having multiple versions of it and
running it on different OS and location without any service
interruption. It actually treat heterogeneous environments as
• Better mobility : auto-switching to the closest available micro-
cloud increases the “things” mobility.
• Lower costs : since the “VCT” resides on multi-tenants
environments, the whole material change is not more required,
which makes the “things” costs less.
• Integration : since the “VCT” is cloud based, it can easily be
integrated with other VCTs or with other ACT offerings.
VCT 1 VCT 1 VCT 1
Cloud is considered by IoT systems as a partner that provide
storage and workflow management. The objective of the proposed
approach is to extend the cloud services layer for a better
coverage of IoT devices and services.
Note: The proposed approach is not a fogging architecture that
aims to push the computing to the edge. Contrariwise, it aims to
pull the computation from the physical devices to the cloud.
SaaS / ACT
Data, Application Code
OaaS / OCT
Data, Application Code
Management and orchestration of
nodes, cluster Highly availablily,
resilience and fault-tolerence.
Compute, storage and networking
Communication and composition Protocol using VCT and CVCT descriptions