A Machine Learning Model for Detection of Docker-based APP Overbooking on Kubernetes

Flashcards on Docker-based APP Overbooking on Kubernetes

Resource allocation overbooking

Allocates more virtual resources than available on physical hardware, potentially degrading service quality.

Docker in cloud computing environments

Increasingly used due to fast provisioning and deployment, but the impact of resource overbooking remains overlooked.

Machine learning model for overbooking detection

Continuously monitors container OS usage and application performance metrics to detect multi-tenancy interference.

Infrastructure-as-a-Service (IaaS)

A service model where the cloud client acquires a configurable virtual machine (VM) according to its service needs.

Service containerization

Lightweight multi-tenancy structure that shares host OS libraries using a container engine (e.g., Docker) to create isolated spaces.

Overbooking

Allocating more virtual resources than the physical hardware can handle, leveraging idle virtual resources.

Docker engine

Ensures isolation between containers and the host OS through namespaces in Linux.

Kubernetes

Framework that manages the scheduling and deployment of containers in a physical cluster.

Problem Statement

Impact of multi-tenancy on dockerized applications performance must be considered simultaneously with migrated services in IaaS clouds

Machine learning model steps

Composed of Containerized Monitoring and SLI Deviation Detection to monitor application and container OS metrics.

Containerized Monitoring module

Continuously monitors containerized OS and app performance metrics within the docker environment.

APP Collector

Collects application performance metrics within the docker environment.

OS Collector

Collects containerized OS usage metrics.

SLI Deviation Detection

Identifies multi-tenancy issues as a supervised machine learning classification task.

Proposed model

Detects resource overbooking within the client domain with high detection accuracy.