Page 63 - Proceedings of the 2018 ITU Kaleidoscope
P. 63
Machine learning for a 5G future
4.2 Emergency Healthcare System interoperability that makes it easier to monitor the physical
world, send instructions to them and collect data, move the
AI Microservices in the Fog-Clouds could be used to data across the Fog up to the Cloud where it can be stored,
provide smart healthcare services where real-time vital sign aggregated, analyzed and turned into information that can
and other data must be processed, and an instant decision be acted upon. One important aspect of this platform that
has to be communicated to healthcare providers such as relates to AIMS is its capability that allows data to travel
emergency health services. A microservice, might be northwards and laterally to other edge gateways, or back to
responsible for collecting the data, another for processing devices, sensors and actuators. However, the edge gateways
the data and other for predicting, and another for deciding only function as data collectors or aggregators for the IoT
on action to take based on the prediction. The AI algorithms devices from which such data is transmitted to the cloud
that these microservices implement are deployed on cross- data centers. The EdgeXFoundry platform does not provide
border Fog-Cloud systems. This kind of system would integration with other cross-border Cloud-IoT platforms
provide low-latency, privacy, trust and secured mobility and also does not incorporate intelligence at the edge of the
and location-aware supports to the individuals in the 5G network to allow application of AI algorithms for data
environment. processing and analyzing IoT data for intelligent decision
making. Another is the MUSA project sponsored by the
To facilitate AI-powered 5G applications such as the European Union [17]. MUSA is a distributed multi-cloud
security surveillance and emergency healthcare as shown in application platform over heterogeneous cloud resources.
Figure 3, AIMS infrastructure hierarchically incorporates Its components are deployed in different cloud service
Cloud and Edge computing with AI and 5G technologies. providers and work in an integrated way and transparently
AIMS provides multi-level AI components located from the for the end users. BigClouT [18] is another similar ongoing
Smart Edges (ROOF/Fog) of things to the Cloud centers. project sponsored by the European Union that leverages the
Thus, the AIMS enables various levels of intelligence, power of Cloud computing, IoT and Big data analytics to
which are deployed at ROOF/Fog/Cloud layers, to be provide distributed intelligence in a smart city network. The
developed as independently deployable microservices AIMS aims to define and develop an integrated platform
(AIMS components). These AIMS components can then be architecture for the incorporation of multi-clouds systems
incorporated based on message driven communications and IoT for AI based services.
provided by the platform, allowing easier extensibility,
interoperability, evolution, integration and composition of 5.2 Specifying essential components and interfaces to
high-level, complex AI-powered 5G services. support data-driven AI services
5. CHALLENGES The AIMS infrastructure consists of broad variety of
heterogeneous nodes, devices, protocols, etc. That interacts
The 5G integrated AIMS platform is envisioned to address in diverse operating conditions from ROOF to the Cloud.
important challenges of an advanced and efficient federated This heterogeneity raises important question of how
cloud platform with IoT for AI applications. It will be microservices deployed across this ecosystem of the
designed to offer distributed AI services (as a microservice) federated AIMS platform would be able to communicate to
over 5G networks, leveraging multi-cloud computing, IoT exchange information and data that are in different formats.
and Big Data technologies. The popular solution would be to design a unified
middleware framework, providing the abstractions of
5.1 Defining an integrated platform architecture for various layers on top of AIMS to hide this complexity from
Cloud, AI and 5G the microservices and allow them to fluidly exchange not
only heterogeneous data and information but also
One of the key challenges is how to define an integrated intelligence seamlessly. Thus, various components and
reference architecture for multi-cloud IoT based interfaces for communication across a federation of ROOF,
microservices, enabling intelligent data acquisition and Fog and Cloud platform would be specified. This
analysis through integrated protocols and standards with middleware and its associated interfaces should be designed
uniform access while supporting different interactions to guarantee interoperability between the federated ROOF,
between various IoT services deployed on federated cloud Fog and Cloud elements, coordinating the life cycle of the
systems at the 5G networks. Presently, there are whole tasks of various microservices taking part in
frameworks providing solutions in this direction. A good delivering intelligence as a service. Components for
example is the EdgeXFoundry open source platform communication, configuration, microservice and resources
developed for the edge of the network [16]. It interacts with discoveries, composition via orchestration or choreography
the physical everyday working world of devices, sensors, and other related service interfaces would be specified and
actuators and other IoT objects. It has been designed as a designed.
framework for industrial IoT edge computing, enabling
rapidly growing community of IoT solution providers to 5.3 Supporting the harmonious management of
work together in an interoperable ecosystem of components computing resources
to reduce uncertainty, accelerate time to market and
facilitate scale. This platform brings the much-needed
– 47 –