Page 59 - Proceedings of the 2018 ITU Kaleidoscope

AI AS A MICROSERVICE (AIMS) OVER 5G NETWORKS

1
2
Gyu Myoung Lee , Tai-Won Um , Jun Kyun Choi 3
1
Liverpool John Moores University (LJMU)
2
Chosun University, Corresponding author
3
Korea Advanced Institute of Science & Technology (KAIST)

ABSTRACT A microservice is a software development technique that
structures an application as a collection of loosely coupled
As data-driven decision-making services are being infused services. Decomposing an application into different smaller
into Internet of Things (IoT) applications, especially at the services can improve modularity and make the application
5G networks, Artificial Intelligence (AI) algorithms such as easier to develop and deploy [2].
deep learning, reinforcement learning, etc. are being
deployed as monolithic application services for In this article, we propose a new concept of AI as a
autonomous decision processes based on data from IoT microservice (AIMS) as an IoT data-driven intelligence-
devices. However, for latency sensitive IoT applications provisioning infrastructure with the 5G capability to
such as health-monitoring or emergency-response provide intelligent connectivity as services closer to the
applications, it is inefficient to transmit data to the Cloud Things by leveraging resources of emerging computing
data centers for storage and AI based processing. In this technologies like Real-time Onsite Operations Facilitation
article, 5G integrated architecture for intelligent IoT based (ROOF) [3], Fog [4] and Edge computing. The AI services
on the concepts of AI as a microservice (AIMS) is presented. will be provided as microservices, implementing
The architecture has been conceived to support the design lightweight functions that have been factored from the AI
and development of AI microservices, which can be algorithms or processes. We envision the proposed system
deployed on federated and integrated 5G networks slices to to allow AI functionality to be infused into 5G networks as
provide autonomous units of intelligence at the Edge of distributed, composable microservices consisting of
Things, as opposed to the current monolithic IoT-Cloud independent virtual components that can be deployed on the
services. The proposed 5G based AI system is envisioned as federated ROOF-Fog-Cloud continuum to improve
a platform for effective deployment of scalable, robust, and scalability, interoperability and cutting down latency for
intelligent cross-border IoT applications to provide real-time 5G applications. Thus, the proposed 5G integrated
improved quality of experience in scenarios where real- platform allows AI services to be provided seamlessly not
time processing, ultra-low latency and intelligence are key only at the centralized data centers but essentially, also at
requirements. Finally, we highlight some challenges to give the edge, closer to the IoT devices. The system and its
future research directions. concepts advocate an architectural principle based on the
abstraction that provides end-to-end fabrics for composing,
Keywords – AIMS, Microservice, AI, 5G provisioning, deploying, managing and monitoring AI
services regardless of whether such services are composed
1. INTRODUCTION from the ROOF, Fog or cloud microservices. The proposed
5G architecture allows hierarchical and horizontal
Over the last decade, the number of Internet-enabled federation of smaller data centers from the Edge to the
devices has run into billions, in fact exceeding the number Cloud data centers continuum so that AI features and
of people on the planet. The emergence of these devices functions can be incrementally composed from
and the large volume of data they generate have a microservices in such hierarchy as dictated by the required
significant impact on our day-to-day existence in diverse and available networking and computing resources at the
ways. Thus, with this development and the success of ROOF and Fog levels to independently execute such AI
Cloud computing, a whole new networking paradigm has functions. In fact, this architectural design would allow
emerged known as the fifth generation wireless networks cross-border microservices to be composed from different
(5G), leading to the development of completely new local ROOF and Fog microservices over 5G network slices.
platform for large scale distributed applications and mobile
services in variety of application domains such as the Smart The remainder of this article is organized as follows.
City, Smart Grid, Healthcare Monitoring, intelligent Background information on 5G networks and distributed
transport, Smart manufacturing, etc., exploiting the Cloud is provided in Section 2. The proposed AIMS
humongous data generated by the “smart” objects [1]. infrastructure is described in Section 3. Section 4 shows
AIMS use cases. In Section 5, we highlight some

978-92-61-26921-0/CFP1868P-ART @ 2018 ITU – 43 – Kaleidoscope