Internet of Things Architecture

rollback

functionality (DC

PS.30)

Use

asynchronous

processing

Asynchronous-

aware functional

component (DC

PS.35)

No impact No impact

Table 21 : Tactics and corresponding Design Choices for Performance and Scalability.

Replication

The functional components (DC PS.1) and the information (DC PS.2) stored
can be replicated to increase performance and scalability (DC PS.3). Having a
single functional component is often against good scalability. The availability of
information depends on the availability of the IoT Device. Having instances of
functional components and information available remotely (for example, in the
cloud) usually increases both scalability and performance (DC PS.4).
Nonethless, in this case one needs to be enough connectivity and bandwidth
provided.

Prioritize Processing

To be able to prioritize processing the functional components needs to be
aware that it might be required to prefer one type of processing over the other.
Therefore, the information model needs to be able to provide information that
indicates priorities of processes, for instance high, normal, or low.. In terms of
deployment the prioritized processing can be done with the help of the network
stack (DC PS.7) or there can be different functional components for the different
priorities (DC PS.8)

Partition and Parallelize

Partition and Parallelize aims towards increase both scalability, as well as,
performance by making the functional components aware of multi-
threading/multi-programming (DC PS.9). Furthermore the information needs to
be partitionable (reduce interdependencies between information) (DC PS.10).
The deployment can help a lot in partitioning, as in IoT access to IoT services
are often locally distributed. This can be done either location aware (DC PS.11),
or based on a data-flow model (DC PS.12).

As an example, the Virtual Entity resolution could be location-oriented, where a
resolution server (RS) is responsible for indexing all connected things in a
certain geographical area, called indexing scope. A Catalogue server then
creates the Catalogue Index of every RS‘ indexing scope. A resolution request
is redirected towards the RS whose indexing scope intersects the search scope
of the request. Large-scale IoT systems are expected to have multiple
administrative domains that must be handled by a federated resolution
infrastructure. Different domains interact with each other by the means of a
central domain directory or domain catalogue. Another possibility would be a