A matrix approach to performance data modeling, analysis and visualization

For large-scale systems depending on numerous parameters and variables, matrices are considered to be a very useful notation. Techniques of linear algebra can be used to analyze the system and understand its behavior. To apply the matrix notation to large-scale computer systems is a novel approach to gain insight into its concurrent activities, which is not an easy task without special analysis methods. Therefore, we introduce a matrix notation for the performance data. Using a metric-based performance matrix, the transformations to the different analysis domains (e.g. spatial and temporal) are explained. This approach towards performance analysis benefits from the methods of linear algebra and provides a mathematical means of analysis as an alternative to approaches based on database systems