Resource requirements of cloud-hosted applications can change dynamically during application execution. Such dynamic changes are usually handled by automatic scaling mechanisms offered by cloud platforms. The reactive nature of auto scaling introduces a lag between when the need to provision additional resources to an application is detected and the actual allocation of the additional resources. Over-provisioning and manual allocation approaches make suboptimal use of resources. This disclosure describes a machine learning model that utilizes application characteristics and infrastructure attributes to predict the time at which resource requirements are likely to change and to generate the prediction timely such that proactive resource allocation is within the operational capabilities of the infrastructure. Inputs to the model include features based on application and infrastructure attributes. The model can be integrated on any cloud platform or other shared computing infrastructure. Automated dynamic optimization of resource allocation can improve the operational efficiency of the computing infrastructure and reduce operating expenses.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.