Smart network infrastructure: Most big data
applications for smart cities require to have smart
networks connecting their components including
residents’ equipment such as cars, smart house
devices, and smart phones. This network should be
capable of efficiently transferring collected data from
their sources to where big data is collected, stored,
and processed and to transfer responses back to the
different entities that need them in the smart city.
The quality of service (QoS) support in the network
is extremely important for real-time big data applications for smart cities. In these applications, all
current distributed application events should be
transferred in real-time to where they can be processed. These events can be transferred from their
sources as raw events or as filtered or aggregated
events. All generated current row, filtered, and aggregated events can be transferred to a centralized
processing point or to distributed intermediate processing points in the smart network for preprocessing or for further filtering and aggregation
before being transferred to the main decision making unit. The centralized approach is good if the
current generated events are not huge and there are
no limitations on the network resources used to
transfer these events. The distributed approach is
more suitable for huge events such that it is inefficient and sometimes impossible to transfer all the
generated events to a single location within acceptable performance and time bounds. Filtering and aggregation will become important in this case
especially for smart cities as it can help reduce the
amount of generated network traffic and speed up
data processing. This can be done at the event
sources and the intermediate points using an openloop or a closed-loop approach. In open-loop approach filtering and aggregation policies are predefined while in closed-loop approach filtering and
aggregation policies are interactively defined based
on the current events and decisions, current system
and network resources, or external smart city application policies. In both approaches, event filtering
and aggregation should be done without compromising the integrity, accuracy and correctness of the
data being aggregated. This is important to preserve
the quality of the decision making process in the
real-time big-data applications [19].
Advanced Algorithms: Standard algorithms used in
regular applications may not be sufficient or efficient
enough to handle big data applications due to their
unique requirements and pressing need for high
volume high speed processing. For example, most
available data mining algorithms are not very
suitable for big data mining applications as their
design is based on limited and well defined data sets
[33]. Big data applications for smart cities will need
to implement advanced and more sophisticated
algorithms to deal with big data efficiently. Some of
these algorithms need to be designed for real-time
application support while others can be designed for
batch or offline processing. These algorithms need
to be optimized to handle high data volumes, large
variety of data types, time constraints on decision
making processes, and distributed components
across various geographical locations. In addition,
these algorithms need to work effectively across
heterogeneous environments and be capable of
managing and operating in highly dynamic
environments.