Browse Prior Art Database

Adoption of regression model to answer performance questions

IP.com Disclosure Number: IPCOM000243471D
Publication Date: 2015-Sep-24
Document File: 2 page(s) / 64K

Publishing Venue

The IP.com Prior Art Database

Abstract

After product deployment customer reports many questions regarding product performance. They mainly ask if their product / environment is properly configured. In other words how should performance look like of particular feature taking into consideration current settings. For example: our product is doing scan of subnets, storing information about computers and applications. The customers ask how long such scan should take? They also provide the information how long it takes in their environment and want to know if it is OK. Because of many different environment configurations (product and scanned end points) it is impossible to provide such answer. The method we propose below is able to predict feature speed taking into consideration customer's settings.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 63% of the total text.

Page 01 of 2

Adoption of regression model to answer performance questions

The core idea is to use regression models to predict performance metric related to customer configuration and then compare it to current performance metric (reported by customer). If the values are similar that means that product performs correctly. If not additional actions are required (tuning, opening a defect).

1. Create learning set consisting of 4 group of data (Figure 1).

Host benchmarking data - information about performance of hardware (similar to

windows experience index)

Product configuration data - information about configuration parameters - for example; numbers of machines to scan
Host config data - information about machine configuration hw/sw - for example: number of CPU, number of RAM and so on
Product performance data - information about performance metrics - for example: time of scanning particular number of endpoint, response time of UI and so on

Note: such data can be gathered by product agent and automatically attached to defects reported by customer or just send for further analysis. Last but not least, data must be gathered from non-performance defects (to eliminate incorrect Product performance data)


2. Based on learning data create regression model able to predict performance data (Figure 2)

    Note: Regression model is known statistical method: "Most commonly, regression analysis estimates the conditional expectation of the dependent variable given the independent variables - that...