Measuring system normality

摘要: A comparative analysis of transaction time-series is made, for light to moderately loaded hosts, motivated by the problem anomaly detection in computers. Criteria measuring statistical state hosts are examined. Applying a scaling transformation measured data, it found that distribution fluctuations about mean closely approximated steady-state, maximum-entropy distribution, modulated periodic variation. The shape under these conditions, depends on dimensionless ratio daily/weekly periodicity and correlation length data. These values persistent or even invariant. We investigate limits conclusions, how they might be applied detection.