Browse Prior Art Database

Method to consolidate workload goal achievement across groups using a non-satisfaction index Disclosure Number: IPCOM000198298D
Publication Date: 2010-Aug-03
Document File: 3 page(s) / 29K

Publishing Venue

The Prior Art Database


In the IT industry a growing trend towards a higher degree of virtualization can be observed. Virtualized infrastructures (IaaS) are by far the most popular offerings that can be selected from a compute cloud. So, the number of workloads running on virtualized infrastructures consisting of one or more virtual servers is expected to grow as well.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 36% of the total text.

Page 1 of 3

Method to consolidate workload goal achievement across groups using a non-satisfaction index

1 Introduction
1.1 Motivation

In a workload management system, like the z/OS Workload Manager, or the Platform Performance Manager coming with the new generation of System z (IBM zEnterprise System), workloads can be managed towards goals that can be expressed close to business terms rather than in resource terms. It is the workload management component's task to translate these business goals into assignments and/or needs for particular resources, such as CPU, memory, or I/O capacity. For example, if a response time goal is not met, the workload management system understands what the current resource assignments are for any given workload and what additional resources would be required such that the goal is met. Of course, taking away resources from one workload in favour of another workload has to be done according to certain rules. In the z/OS Workload Manager, for example, the Importance of any piece of work is used to prioritize the list of possible receivers. Another criteria for prioritization that such a workload management component needs is a measure of how far the goal is missed or exceeded. Understanding this allows to project the result of a resource re-assignment and to decide whether the re-assignment will happen or not. In the z/OS Workload Manager, this metric is called Performance Index (PI).

With the growing number of workloads that are managed in such a way, the need for a consolidated view is raised as well. For example, the new IBM zEnterprise System will allow you to combine up to eight physical boxes into a so-called Ensemble in which several hundreds, maybe thousands of virtual servers might be running. Now consider that you could have 50 or 100 different workloads in multiple such Ensembles, how can one quickly decide which workload requires attention first, if goals are not met? So, the consolidation should not simply show workloads missing goals but it should also actually account for the importance of that workload.

1.2 Objective

The method described below, proposes a method to compute a non-satisfaction index that expresses the degree of dis-satisfaction of a workload and that can be aggregated across workloads to yield a single index that could be presented, for example, on a dashboard. Looking at the dashboard, a supervisor can immediately tell which area in the enterprise requires attention first and from there he can navigate quickly to the workload in trouble.

2 Solution
2.1 Description

The proposed method to create the non-satisfaction index makes use of the knowledge that a workload management component has when deciding what workload is in most need of resources. As mentioned above, there are two measurements, the Importance and the PI. The workloads in most need of resources are ideally be sorted by decreasing Importance and by decreasing PI. Importances are scalar values, for example from 1 to 5 with 1...