TELEMETRY-DRIVEN FINOPS AUTOMATION FOR COST OPTIMIZATION IN MULTI-CLOUD ENVIRONMENTS

Abstract

Abstract: The dynamic nature of the provisioning and instances mixed with the heterogeneous configurations and cost-based pricing models that have come with the rapid implementation of multi-clouds has presented a tremendous challenge to cost governance.   The constant over-provisioning, idle computer resources and the ineffective allocation of storage are often the cause of unnecessary operational cost.   In this study, a telemetry-based FinOps automation framework will be proposed gaining insights on structural and behavioural inefficiencies across the Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform (GCP).   Descriptive statistics, correlation modelling, a threshold-based idle detection and a rule-based rightsizing decision framework were used to analyse a production-representative multi-cloud data.   The cross-provider cost-efficiency analysis has been based on the standardised metrics, such as cost per vCPU-horn.   Findings indicate a moderate on the whole utilisation, heavy reliance on cost in dependence on provided capacity rather than on the actual requirement to work and high potentials of optimisation.   The model of simulation of an optimisation impact simulation proves the cost savings that can be measured by the automated policy of idle shutdown and rightsizing.   The results indicate that inefficiencies in clouds are mainly due to provisioning methods than provider choice which highlights the necessity of telemetry constantly and automated government.   The suggested structure involves the provision of a repeatable and systematic approach to enhancing cost-performance fit in heterogeneous multi-cloud setup and promoting the active implementation of FinOps.