3 Engineering IoT Networks
99
tolerates high and variable delay, but for closed-loop control applications and alarm
notification, the delay should be kept small and constant. The error rate is usually
not negligible in case of wireless channels, and this aspect should be considered in
the deployment of the application. Network standards for the Edge Network will be
presented in the main body of this chapter.
The Edge Network is connected to the Internet through nodes acting as gateways
between the two worlds. Internet architecture and protocols are not the topics of this
book even if some of them will be found in the next sections since they are also used
in the Edge Network. The Internet moves data to/from user’s computer (e.g., for data
visualization) and data centers for storage as well as processing (e.g., with machine
learning algorithms); the use of data centers is related to the concept of Cloud [1].
Gateways play an important role in the IoT scenario of Fig. 3.1. Their basic function
consists in moving data between Edge Network protocols and Internet protocols.
Since they are more standard computational devices and are usually powered by
stable sources, they can also perform data processing, thus moving “intelligence”
from the cloud closer to the “things.” This technique may reduce response delay
and is related to the concept of Fog Computing or Edge Computing [2].
This chapter focuses on the red part of Fig. 3.1. It has some distinguishing
features:
• strict dependence between application and communication aspects;
• system-of-systems nature;
• strict relationship with the environment.
IoT applications feature a strict dependence between application requirements
as well as constraints and communication aspects. While in the traditional Internet,
communication requirements are quite uniform for the users, in IoT applications
there is a large variability of requirements for data rates, delay, and error rates. For
instance, agricultural sensor and gas meters produce meager data rates, while videosurveillance cameras generate a large amount of data. Furthermore, sensors can be
very far from the power grid, and, therefore, the corresponding protocols should
consume a small amount of energy to allow long autonomy with batteries or the
use of environmental energy sources. In factory automation, level and variability of
delays are crucial for the correct behavior of closed-loop control algorithms.
Many IoT applications can be regarded as a system-of-systems since even if the
various nodes can independently operate, they interact together to achieve the good
behavior of the global application [3]. For instance, in a building automation application, the final objective may be to achieve reasonable control of the temperature,
and it does not matter the set of nodes that provides such functionality, as long as
the global application behavior satisfies design objectives. Thus, these applications
pose new questions to designers, traditionally mainly interested in the specification
of each single network node as done for Internet servers and clients. Most relevant
issues are:
• finding the optimal number of nodes to achieve the common mission;
99
tolerates high and variable delay, but for closed-loop control applications and alarm
notification, the delay should be kept small and constant. The error rate is usually
not negligible in case of wireless channels, and this aspect should be considered in
the deployment of the application. Network standards for the Edge Network will be
presented in the main body of this chapter.
The Edge Network is connected to the Internet through nodes acting as gateways
between the two worlds. Internet architecture and protocols are not the topics of this
book even if some of them will be found in the next sections since they are also used
in the Edge Network. The Internet moves data to/from user’s computer (e.g., for data
visualization) and data centers for storage as well as processing (e.g., with machine
learning algorithms); the use of data centers is related to the concept of Cloud [1].
Gateways play an important role in the IoT scenario of Fig. 3.1. Their basic function
consists in moving data between Edge Network protocols and Internet protocols.
Since they are more standard computational devices and are usually powered by
stable sources, they can also perform data processing, thus moving “intelligence”
from the cloud closer to the “things.” This technique may reduce response delay
and is related to the concept of Fog Computing or Edge Computing [2].
This chapter focuses on the red part of Fig. 3.1. It has some distinguishing
features:
• strict dependence between application and communication aspects;
• system-of-systems nature;
• strict relationship with the environment.
IoT applications feature a strict dependence between application requirements
as well as constraints and communication aspects. While in the traditional Internet,
communication requirements are quite uniform for the users, in IoT applications
there is a large variability of requirements for data rates, delay, and error rates. For
instance, agricultural sensor and gas meters produce meager data rates, while videosurveillance cameras generate a large amount of data. Furthermore, sensors can be
very far from the power grid, and, therefore, the corresponding protocols should
consume a small amount of energy to allow long autonomy with batteries or the
use of environmental energy sources. In factory automation, level and variability of
delays are crucial for the correct behavior of closed-loop control algorithms.
Many IoT applications can be regarded as a system-of-systems since even if the
various nodes can independently operate, they interact together to achieve the good
behavior of the global application [3]. For instance, in a building automation application, the final objective may be to achieve reasonable control of the temperature,
and it does not matter the set of nodes that provides such functionality, as long as
the global application behavior satisfies design objectives. Thus, these applications
pose new questions to designers, traditionally mainly interested in the specification
of each single network node as done for Internet servers and clients. Most relevant
issues are:
• finding the optimal number of nodes to achieve the common mission;