Feature: Right sizing of Kubernetes resources with ML models

[satyampsoni]

Description

To ensure that a “resource-hogging” situation does not occur, the best practice for running application workloads on Kubernetes is to set a fixed number of resource requests and limits within it. However, it can be difficult to accurately estimate the correct resource allocation constraints. One of the possible solutions is to train an ML model to analyze historical data from an external source such as Prometheus and recommend the right sizing for the resources that an application should be allocated based on its usage pattern.

Deliverable

• Extract and preprocess historical resource usage data from Prometheus.
• Develop, train, and optimize an ML model for resource allocation recommendations.
• Validate recommendation accuracy against real-world usage.
• Benchmark improvements over static resource limits.
• Provide setup, deployment, and usage guides.

Key Competencies:

• Basic knowledge of programming languages
• A willingness and eagerness to learn new and implement new technologies

Recommended Skills to have:

• Go
• Prometheus
• Kubernetes

Mentors: Nishant Kumar, Kartik Singhal, Prakhar Katiyar Skill Level: Medium Time: ~175 hrs