Browse Prior Art Database

Goal-Oriented Volume Selection Using the IBM Common Information Model for the Enterprise Storage Server

IP.com Disclosure Number: IPCOM000126402D
Original Publication Date: 2005-Jul-15
Included in the Prior Art Database: 2005-Jul-15
Document File: 6 page(s) / 100K

Publishing Venue

IBM

Abstract

This article presents a novel algorithm that performs volume selection by representing volumes as points in an n-dimensional feature space, and minimizing distances between these points as a selection criterion. This algorithm is proposed in the context of the IBM Enterprise Storage Server*, but can be extended for other enterprise subsystems as well.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 34% of the total text.

Page 1 of 6

Goal-Oriented Volume Selection Using the IBM Common Information Model for the Enterprise Storage Server

Overview

Disclosed in this article is an algorithm that performs volume selection by representing volumes as points in an n-dimensional feature space, and minimizing distances between these points as a selection criterion. The author proposes that the distance between any two points in the feature space corresponds to quantifiable "similarity" between the two points. The selection algorithm is biased to select target volumes that do not correspond to the same physical resource as the source volume, with the goal of minimizing performance degradation. In the proposed algorithm, similarity is quantified by the minimization of an objective function. The objective function selected for use with the selection algorithm is minimization of the Euclidean distance between the performance profiles (to be defined) of the source and target volumes.

General Performance Considerations

In many instances, the way data is allocated within the back-end storage can greatly affect the performance of the end application. There has been a great deal of study devoted to storage provisioning, volume selection, and optimal data layout. While the author does not imply the target selection algorithm to be optimal, a biasing heuristic has been incorporated into the algorithm, with the goal of minimizing possible performance degradation due to a selection decision. The heuristic imposed by the volume selection algorithm for the IBM Enterprise Storage Server (ESS) is, if possible, select a target volume that resides on a different ESS rank than the source volume, in order to minimize performance degradation due to resource contention.

Performance Profiles

Performance profiles are a vital component of the selection algorithm. Conceptually speaking, performance profiles can be considered as points in an n-dimensional performance feature space, where n is the number of performance metrics used in characterizing the performance of a volume. To better illustrate this conceptual view, let n = 2. We now have a two-dimensional feature space, i.e., the performance profiles of the source and target volumes consist of only two performance metrics. In this case, the format of the performance profile of the source and target volume would be of the form S:<x1,x2 > and T:<x1,x2 >, respectively, where xi is a performance metric collected by IBM Common Information Model (CIM) for ESS. More on how these performance metrics are obtained will be given later. A diagram of a two-dimensional feature space is given below. Here each point corresponds to a volume, given by plotting a volume's performance in the performance space.

1

Page 2 of 6

Figure 1. Two-dimensional performance feature space.

While clearly conceptual, the diagram is useful in illustrating how the selection algorithm works. The selection algorithm accepts a source volume (for purposes of illustration, a random poin...