Browse Prior Art Database

Statistical Correlation of Hypervisor Performance Data

IP.com Disclosure Number: IPCOM000246749D
Publication Date: 2016-Jun-29
Document File: 2 page(s) / 54K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a process for calculating the Pearson product-moment correlation coefficient between a user-selected hypervisor performance metric and all other available hypervisor performance metrics, and then displaying all of the correlations. This process facilitates the calculation of the coefficients for a given hypervisor performance variable and reveals the hypervisor performance variables that are most strongly correlated with it.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 2

Statistical Correlation of Hypervisor Performance Data

When a hypervisor (and the hosting physical machine) experiences poor performance, locating the root cause of this degradation in performance is difficult. Many operating systems and hypervisors capture performance data. Usually, this data is in record form

with performance metrics displayed at the time intervals. This data reveals an unexpected result, such as an increase in response time for a virtual machine at a certain time, but it is not always easy to determine other metrics related to this performance degradation.

The novel contribution is a process for calculating the Pearson product-moment correlation coefficient between a user-selected hypervisor performance metric and all other available hypervisor performance metrics, and then displaying all of the correlations. This process facilitates the calculation of the coefficients for a given hypervisor performance variable and reveals the hypervisor performance variables that are most strongly correlated with it. The outcome is a clearer picture of which variables are influential to a given performance issue. The root performance problem correlates to the symptoms.

Performance data for hypervisors are often found in record form, where performance

data is displayed for each time interval. This can be state sampled data; the state of the system is captured at a given point at regular intervals (e.g., every 30 seconds or one minute), event or configuration data, counter data where metrics are counted by the system, or hardware data.

The first step is to parse this data into a format that is appropriate for performing correlation calculations. The performance data records may not all be in a uniform format; therefore, the parser program must be programmed such that a...