Page 64 - Proceedings of the 2018 ITU Kaleidoscope
P. 64
2018 ITU Kaleidoscope Academic Conference
In the proposed AIMS integrated platform, large determine based on the available resources, what level of
heterogeneous and distributed IoT devices will produce intelligence should be provided by a microservice, what
huge volume of data at rapid velocity. Gleaning meaningful tasks or functions should be executed and at what layer of
information and insights from this data using distributed AI the infrastructure are important technical challenges.
services will require a shift from the traditional
architectural style to a more agile approach that allows 5.5 Supporting trusted AI services
more robust scalability, evolvability and maintainability of
large-scale distributed multi-cloud IoT systems. The AIMS based applications will process large volume of
Microservice architecture, as one of the recent trends in the data using distributed microservices from ROOF to Cloud
design and development of agile distributed systems, continuum of the platform. Thus, as distributed and
defines a new approach to designing and developing a interoperating microservices execute intelligence based on
single application as a suite of smaller services, each data from the physical devices, such data can be
running in its own process and communicating with compromised as they may be exposed to malicious third
lightweight mechanism to execute just one task. Such parties. In fact, malicious microservices can be injected into
services are small, highly decoupled, independently the system to wreak havoc or to provide false or misleading
deployable, focusing on doing a small specific and decisions. This is a crosscutting challenge since it does not
interdependent task that can provide some level of only affect a layer but all layers and aspects of the AIMS
intelligence and yet when combined with other tasks ecosystem, from radio communications to the microservices
provide higher or deeper intelligence depending on across the 5G networks. Additionally, the interfaces
available and required resources. In order to achieve a between Cloud, Fog and ROOF computing are potential
much bigger task, these services can be combined to realize sources of vulnerability and consequently may lead to
such functionality. One of the key challenges therefore that corruption of IoT data and services. To ensure trust, privacy
need to be addressed is that of dynamic allocation and and security, capabilities for end-to-end encryptions,
orchestration of resources for the distributed microservices intrusion detection and prevention of unauthorized
in the AIMS federated platform depending on what microservices or services will be required. Trust
compute resources are currently available and how much of management should be investigated as a useful technology
resources are required by the current microservices for the for providing such required security services. How can trust
task execution. Although, there has been existing resource management be used to provide security, dependability and
allocation in 5G networks, however, there is no yet concrete reliability for AIMS and associated data at various layers of
solution for resource allocation for integrated ROOF, Fog the ROOF, Fog and Cloud integrated platform? For users’
and Cloud platform. Even for the more mature Fog, needs and rights to be enforced as autonomous
resource and service orchestration remains a challenging microservices exploit IoT data to infuse intelligence into
research problem. For the AIMS platform, there would be IoT applications, there is need to investigate integrated and
several microservices sharing resources and this might federated ROOF, Fog and Cloud platform to propose the
result in resource contention and interference. Thus, new best and unique trust and security mechanisms for
mechanisms and strategies for dynamic and fluid resource enforcing integrity, dependability and reliability of the
allocation and scheduling would be investigated to reduce platform and its services.
response time for task execution across the 5G integrated
AIMS platform. 6. CONCLUSION
5.4 Applying new mechanisms using intelligence in data This article proposes an IoT data-driven intelligence-
lifecycle provisioning infrastructure with the 5G capabilities to
provide intelligent connectivity as services closer to the
To provide distributed intelligence at the edge of things, a Things by leveraging the compute resources of a
critical factor for deploying AI services on the ROOF-Fog- hierarchically integrated computing environment (ROOF-
Cloud integrated infrastructure is more related to Fog-Cloud). The proposed AIMS aims to provide a
application partitioning or factoring, real-time service lightweight platform for effective deployment of scalable,
composition, data mobility and aggregation. To address robust, and intelligent cross-border 5G applications. We
these issues, there is need for new mechanisms for factoring have envisioned the proposed architectural approaches in
or decomposing AI services into functions that can be terms of system perspectives to allow AI functionality to be
delivered as re-usable microservices for executing specific infused into 5G networks as distributed, composable
smaller tasks. These new mechanisms for service microservices consisting of independent virtual components
composition must be developed to achieve a fluid decision that can be deployed on the federated Roof-Fog-Cloud
making process exploiting raw data from the physical continuum to improve scalability, interoperability and
devices, extracting meaning and insights in order to achieve cutting down latency for real-time 5G applications. In this
the DIKW at the ROOF, Fog and Cloud layers of the article, we have also highlighted some challenges to give
infrastructure’s hierarchy. Such mechanisms should support future research directions.
the dynamic discovery, composition and relocation of
AIMS according to the required and available resources
across the integrated nodes on the AIMS platform. How to
– 48 –