Browse Prior Art Database

Mechanism of machine learning to automate corrective actions in a virtualized environment

IP.com Disclosure Number: IPCOM000246356D
Publication Date: 2016-Jun-02
Document File: 3 page(s) / 36K

Publishing Venue

The IP.com Prior Art Database

Abstract

The utilization metrics of each Virtual Machine is captured through a Performance Monitoring tool. This tool is part of the Server/System Management stack and collects and provides the information of the current and historic utilization metrics of the Virtual Machine. The idea is to develop an algorithm which will enable the tool to self learn the corrective actions required for recovering from performance bottlenecks and critical errors in a Virtual Machine or Server. The learned actions will be persisted/remembered and will be invoked by the the Performance Monitoring tool itself in an automated way in the subsequent occurrence of such alerts where the thresholds have been exceeded.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 34% of the total text.

Page 01 of 3

Mechanism of machine learning to automate corrective actions in a virtualized environment

The utilization metrics of each Virtual Machine is captured through a Performance Monitoring tool. This tool is part of the Server/System Management stack and collects and provides the information of the current and historic utilization metrics of the Virtual Machine. The utilization metrics include information about processor, memory, network and storage metrics. The user has to launch the Graphical User Interface(GUI) of the Performance Monitoring tool to view the utilization details of a Virtual Machine. When the resource utilization crosses a defined threshold, there are possibilities that the user might not be looking at the Performance Monitoring tool GUI. This would lead to instances when there are no actions taken by the System administrators when the utilization metrics exceed optimal thresholds. Another problem with this model is that the System administrator has to intervene every time a threshold is exceeded and perform a configuration change to bring back to the Server to optimum resource utilization. Along with this, the Server management console provides general alerts on system health or virtual machine health. These errors also need to be viewed in the serviceable events GUI on the management console. Now this also holds a similar problem of the system administrator missing the alerts or having to take repetitive action for the same alert.

The current solution requires a manual intervention everytime a performance bottleneck happens or a threshold is surpassed or for any kind of system or virtual machine health alerts. The drawbacks of this solution include constant manual observation, repetitive corrective actions and the possibilities of missing a problem. These drawbacks could cause a Server downtime in a customer environment.

The idea is to develop an algorithm which will enable the GUI tool to self learn the corrective actions required for recovering from Performance bottlenecks and server error in a Virtual

Machine or Server. The Performance Monitoring tool should allow the System administrators to define utilization thresholds at the Virtual Machine level. This threshold settings should be configured by the System administrators based on the workloads deployed in the Virtual

Machine. When the thresholds are exceeded, the Performance Monitoring tool will generate noticeable alerts for the System administrator to take actions. These alerts could be displayed as a notification on the GUI or sent through a message or mail to a configured recipient.

The Performance Monitoring tool and the Management Console tool should provide an option of "Remember Me" along with every Alert. When the System administrator enables this option, then the tool learns the corrective action performed by the System administrator for that specific alert. The tool will also provide an option to repeat the corrective steps performed by the System a...