Resource Optimization with DRL-driven Real Time Service Placement Strategy in Edge-Cloud Continuum

Jeyhun Abbasov
The growth of Internet of Things (IoT) devices and the need for data intensive applications has led to Edge-Fog-Cloud architecture, known as Edge-Cloud continuum. Cloud computing is utilized for handling and keeping the large amounts of data produced by IoT devices. One of the major limitations of Cloud computing is network latency. Due to these limitations, the Fog computing is introduced. Fog computing provides near real-time services and saves network resources. However, Fog computing has lower resource capacity comparing to Cloud computing. Edge computing is an extension of Fog Computing where data is processed closer to the source. In our study, we present a Deep Reinforcement Learning (DRL) real time service distribution solution without compromising the Quality of Services (QoS) in Edge-Cloud continuum. The services are offered by Fog and Cloud environments. The request that is coming from user is sliced in Edge and distributed between the Fog and Cloud environments using DRL. The proposed DRL algorithm is implemented and evaluated in terms of its success rate, distribution of service request slices, and etc. Furthermore, the study delves into the intricate dynamics of three distinct data-intensive applications, revealing insights into their performance and resource utilization within the Edge-Cloud continuum.
Graduation Thesis language
Graduation Thesis type
Master - Computer Science
Chinmaya Kumar Dehury
Defence year