The emerging application space for smart objects requires scalable and
interoperable communication mechanisms that support future innovation
as the application space grows. IP has proven itself a long lived, stable,
and highly scalable communication technology that supports both a wide
range of applications, devices, and underlying communication
technologies. The IP stack is lightweight and runs on tiny, battery
operated embedded devices. IP therefore has all the qualities to make
‘The Internet of Things' a reality, connecting billions of communicating
devices. Internet Protocol for Smart Objects (IPSO) Alliance
SMART OBJECTS are small computers with a
sensor or actuator and a communication device,
embedded in objects such as thermometers, car
engines, light switches, and industry machinery.
Smart objects enable a wide range of applications
in areas such as home automation,
building automation, factory monitoring, smart
cities, health management systems, smart grid
and energy management, and transportation.
Until recently, smart objects were realised
with limited communication capabilities, such
as RFID tags, but the new generation of devices
has bidirectional wireless communication and
sensors that provide real time data such as
temperature, pressure, vibrations, and energy
measurement.
Smart objects can be battery operated, but
not always, and typically have three
components: a CPU (8 , 16 or 32 bit microcontroller),
memory (a few tens of kilobytes) and
a low power wireless communication device
(from a few kilobits/s to a few hundreds of
kilobits/s). The size is small and the price is
low: a few square millimetres and few dollars.
The technical development in low cost sensors
and actuators combined with low power
communication technologies such as IEEE
802.15.4, low power Wi-Fi, and power line
communication has been rapid. Nevertheless,
the emergence of smart object applications has
not been as fast because the large number of
proprietary or semi closed systems has lead to
partial and non-interoperable solutions.
The current situation for smart objects is
similar to what computer networks looked like
about two decades ago: islands of computers
communicating with their own protocol, for
example SNA, IPX, and Vines, interconnected by
complex multi-protocol gateways. Subsequently,
these architectures evolved to IP based
tunnelling mechanisms such as DLSw or XOT.
Today, these networks operate on fully end to
end IP based architectures.
Many of today's non IP based sensor architectures
are evolving toward a protocol
translation gateway model, similar to the path
computer networks went through before quickly
moving to fully IP based architectures. Have
we not learnt from the past? Protocol gateways
are inherently complex to design, manage, and
deploy. The network fragmentation leads to
non efficient networks because of inconsistent
routing, QoS, transport and network recovery
techniques. end-to-end IP architectures are
widely accepted as the only alternative to
design scalable and efficient networks of large
numbers of communicating devices.
The Internet of Things
To support the large number of emerging applications
for smart objects, the underlying
networking technology must be inherently
scalable, interoperable, and have a solid standardisation
base to support future innovation
as the application space grows.
IP has proven itself a long lived, stable, and
highly scalable communication technology that
supports both a wide range of application, a
wide range of devices, and a wide range of
underlying communication technologies. The
layered architecture of IP provides a high level
of flexibility and innovation. IP already
supports a plethora of applications, such as
email, the World Wide Web, Internet telephony,
video streaming, and collaborative tools. Over
the past 20 years, IP has evolved to support
new mechanisms for high availability, enhanced
security, support of Quality of Service (QoS),
real time transport, and Virtual Private
Networks (VPNs).
IP has a long history as a communication
mechanism for general purpose PC computers
and network servers. It was therefore long
believed that IP was too heavy weight to run on
highly constrained devices. Several recent
lightweight IP stacks have demonstrated that
they can be designed to meet the requirements
of light footprint devices with a few kilobytes of
RAM and ROM, limited processing power and
severe energy constraints.
IP provides standardised, lightweight, and
platform independent network access to smart
objects and other embedded networked devices.
The use of IP makes devices accessible from
anywhere and from anything; general purpose
PC computers, cell phones, PDAs as well as
database servers and other automated equipment
such as a temperature sensor or a light bulb.
IP runs over virtually any underlying communication
technology, ranging from high speed
wired Ethernet links to low power 802.15.4
radios and 802.11 (Wi-Fi) equipment. For long
haul communication, IP data is readily transported
through encrypted channels over the
global Internet.
Memory efficient implementations of the IP
stack show that IP can successfully work in as
little as a few kilobytes of RAM, and require
less than 10KB of ROM. Figure 1 shows the
memory footprint of five embedded TCP/IP
stacks: the open source IwIP stack from the
Contiki operating system, one commercially
available TinyOS based IPv6 stack, the commercially
available NanoStack, and the open source
lwIP stack. Their footprint is around 10KB,
except for lwIP that is around 20KB.
Fig. 1. Memory footprint for five embedded TCP/IP stacks
For power constrained devices, recent standardisation
work has made IP power efficient
enough to run over sub-milliwatt radio links
such as 802.15.4. Such low power operation
enables years of lifetime on typical AA
batteries, even for multi-hop routing nodes.
IP is Scalable. With the global Internet, IP
has proven itself to be inherently scalable. No
other networking technology has ever been
deployed and tested at such an immense scale
and with such a large number of devices. As
smart objects will connect an even larger
number of devices than that of the existing
Internet, scalability is a primary concern.
The next generation Internet protocol, IPv6,
expands the address space of IP to 2128. Such
a large address space has been said to be
enough to provide every grain of sand on the
planet with an IP address.
IP is End-to-End. IP provides end-to-end
communication between devices, without intermediate
protocol translation gateways. Protocol
gateways are inherently complex to design,
manage, and deploy. The objective of a gateway
is to translate or map between two or more
protocols. Such translation, however, typically
requires significant semantic and functional
translation for the protocols to work together.
Mechanisms on both sides usually differ significantly,
thus requiring the adoption of a least
common denominator approach that leads to
inefficient networks because of inconsistent
routing, QoS, transport and network recovery
techniques. With the end-to-end architecture
of IP, there are no protocol translation
gateways involved.
With the IP end-to-end architecture, there is
no single point of failure. Intermediate routers
may fail, but the end-to-end communication
will chose alternate paths through the network.
In contrast, if a protocol translation gateway
fails, the entire network fails.
In the IP architecture, protocols can change
without affecting the underlying network.
Routers operate independently of the protocols
running over them. In contrast, a protocol
translation gateway needs to be updated every
time a protocol changes, no matter how small
the change.
With the success of today's global Internet,
the end-to-end architecture of IP has proven
itself scalable, stable, and efficient. For the
future Internet of things, scalability, stability,
and efficiency is even more important than
ever. IP therefore is the future proof choice for
the Internet of Things.
Smart objects enable a wide range of applications
that will improve our lives in many
areas such as energy management, healthcare,
and safety. The recent progress in low cost
embedded devices is about to make the
Internet of Things a reality. For this to come
true, we must learn from the lessons of the
past and adopt a flexible, scalable, efficient
and open based networking technology. IP has
proven itself to fulfil these requirements and it
is now a fact that IP can meet the strict
requirements of highly constrained smart object
networks.
Adam Dunkels Ph.D is a senior scientist, Swedish Institute of Computer Science
JP Vasseur is an engineer with Cisco Systems
Further information: Video Contiki uIPv6 Demonstration on Atmel Raven Hardware
Back
