Browse Prior Art Database

SLA based autonomic optimzation of infrastructure

IP.com Disclosure Number: IPCOM000033318D
Original Publication Date: 2004-Dec-06
Included in the Prior Art Database: 2004-Dec-06
Document File: 3 page(s) / 43K

Publishing Venue

IBM

Abstract

In an On-demand utility based model, there is a need to dynamically provision resources by the service provider whenever there is a possibility of a Service Level Agreement (SLA) breach. Descibed is a mechanism for providers of an on-demand hosting model to autonomically predict the potentiality of a service level agreement breach, and take corrective action to prevent such an occurence.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 3

SLA based autonomic optimzation of infrastructure

Disclosed is a tool to use the information gathered by a proactive testing tool as an input to autonomically fine tune the configuration of the hosted infrastructure to meet the service level agreements. In addition, this idea can also take a desired load and service level as an input and, in conjunction with the above mentioned testing tool, autonomically adjust the configuration of the hosted infrastructure in order to meet the specified service levels under the specified loads.

Currently there are stand-alone processes which monitor the infrastructure (can be Utility Modelling Infrastracture(UMI), e-hosting, etc.) to detect potential failures in the system, such as whether the database is down, Message Queueing (MQ) queue is full or reached a certain threshold, network connection exists, and system functions (activity, file system usage, etc.). Integrated with other solutions, it can continously check to see if the resources are running out, and trigger a separate process to allocate/deallocate resources as necessary. The key differentiators are as follows:

This tool would not have to be kept running all the time, unlike the processes that currently do 'monitoring' today. This tool can be initiated by the proactive testing suite, and will leverage results of the proactive testing to perform auto-tuning, once complete it can return to sleep mode until the next request. Thus, saving computing resources and only becomes active as required. Essentially being proactive in adjusting customer-specific computing resources preparing for spikes, and ensuring the service provider (IBM* in the case of UMI or e-hosting) fulfills the Service Level Agreement (SLA) to avoid breach penalties.

This tool is also fundamentally different, in that it tries to optimize resources to meet SLAs, whereas the process does not have any SLA information, and so all it can do increase the currently available resources with the aim of preventing them from running out, reaching a final configuration that may be far from optimal. This is a key because simply working with computing resources does not involve SLA, and these processes would first need to gather customer specific SLA data in order to work.

Current arts use tools to monitor usage of resources and increase (as opposed to optimize) the resource availability when the resources start running out. These existing tools, in their present state, do not have a way of optimizing resource usage due to lack of essential data such as parameters to optimize against and the goals to achieve through such optimization.

Below are other features and advantages

Better track SLA at a more granular level, instead of only guarantee 99.9% up time, service provider can negotiate contract on network access and other operational fields because this tool presents a way to pro...