Architecture for distributed systems that facilitates a cloud-native AIOps implementations

Abstract

ABSTRACT : The most popular approaches that have been used such as DevOps improve application operations thanks to the heyday of containers and CI/CD, however, this still requires human intervention in case of failures in any of the system components, since many of the solutions that have been used so far are limited to specific problems, such as reacting to downed servers and scaling them up. Taking into account that every time the operations of distributed systems become more and more complex due to the large number of components that must run at the same time, and also considering that in many applications, even small unavailability can translate into a strong impact on the reliability of the application, which implies an economic impact for the business, it is necessary that the solutions that are created reduce any type of risk and each time all these operations are more automated. Due to this, AIOps arises which uses artificial intelligence techniques such as machine learning and big data to operate and maintain application infrastructures, reduce the operational complexity of systems and automate IT operations processes. It has been proven that the implementation of this type of solution improves the quality of the systems and reduces the MTTD (Mean time to detect an error) from 10 minutes to 1 minute and the MTTR (Mean time to recovery) can be reduced from 60 minutes to 30 seconds [1], which is a very important advance in this world of operations. Despite this, there is still not so much adoption by most companies due to the challenges involved in implementing them in large projects and the fact that there is no clear path to follow when integrating applications to this type of solution that is emerging. In this research, we are going to propose a holistic architecture that makes it easier for cloud-native distributed systems to integrate with these new solutions.