Browse Prior Art Database

Generating Virtual Performance Software Pattern Profiles

IP.com Disclosure Number: IPCOM000234828D
Publication Date: 2014-Feb-10
Document File: 5 page(s) / 82K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method to enable users of the cloud to generate profiles containing performance related data for instances of a pattern by comparing data between the guest operating system (OS) and hypervisor, and then calculating an effective set of load metrics for the Central Processing Unit (CPU), Random Access Memory (RAM), etc. These profiles are calculated and exposed at the pattern level, and users can easily compare these profiles to identify issues while still adhering to security protocol and not compromising another deployed instance of the pattern.

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

Page 01 of 5

Generating Virtual Performance Software Pattern Profiles

The cloud is a dynamic environment that can vary throughout the day depending on user traffic. A user's experience at the pattern level can vary throughout the day depending on the loads being pushed around a particular instance. In cloud environments, it is difficult for an end user of a pattern instance (i.e., collection of virtual machines coupled with middleware and applications) to know if other users of the same pattern are experiencing similar performance or if the current performance experience
is normal given current cloud conditions. This data is often impossible to access within the end user role due to security or administrative restrictions and difficult to normalize if given the necessary metrics by the guest operating system (OS) and hypervisor.

The disclosed method enables the administrator of a virtual instance of a pattern running in a Platform as a Service (PaaS) cloud to understand how a particular instance is behaving compared to other instances of that pattern at varying periods.

Performance data specific to particular cloud layers may not be exposed or translated into a metric that a user of the cloud can interpret for an effective performance analysis of the specific application pattern in relation to various profiles or other user's activity

with the same pattern over time given current cloud conditions.

A method is herein disclosed for enabling users of the cloud to generate profiles

containing performance related data for instances of a pattern by comparing data between the guest OS and hypervisor, and then calculating an effective set of load metrics for the Central Processing Unit (CPU), Random Access Memory (RAM), etc. These profiles are calculated and exposed at the pattern level, and users can easily compare these profiles to identify issues while still adhering to security protocol and not compromising another instance.

The following provides additional detail regarding the method disclosed herein:

    A user can compare a pattern instance to a previously captured pattern profile to quickly find discrepancies such as:


• The database (DB) node in instance "A" is consuming a lot more CPU (relative to the Application Server nodes) than the DB nodes in other instances of the same pattern


• The CPU load on the three application servers in instance "B" is less balanced (relative to each other) than the application server nodes in other instances of the same pattern


• The application servers in instance "C" are consuming significantly more RAM (in aggregate) than a previously captured profile from another instance of the same pattern

In an embodiment of the present method, users can associate one or more "Pattern Performance Profiles" with a pattern. Pattern performance profiles can be captured

1


Page 02 of 5

from instances of the pattern that are currently running, or can be uploaded (e.g., a service provider could provide "benchmark profiles" wi...