Performance Baseline
Database administrators face the ongoing challenge of effectively planning for the future of their database systems. Performance Analysis's adaptive IntelliProfile engine dynamically analyzes time-sensitive operating metrics to produce a baseline (expected) operating range for the instance. A form of quantitative analysis, this process looks at historical performance data to benchmark current performance, and, moreover, to estimate future performance. Good quantitative analysis requires as much data as is available, and IntelliProfile incorporates data on hardware and software configurations, concurrent users, workload size, instance availability, as well as other day-to-day operating specifics.
The baseline algorithm parses sampled data into pre-defined time units (e.g. day of the week) and distinct episodes (e.g. systematically recurring spikes in CPU consumption). These episodes form the basis for projecting normative instance behavior, allowing for both exception monitoring and performance diagnostics. These dynamically-created statistical ranges help administrators configure high and low metric threshold values. The IntelliProfile engine learns the instance's "signature" behavior characteristics via sophisticated algorithms that identify both cyclical and periodic behavior to minimize and/or neutralize random noise distortion (for example, an employee absent due to illness). The engine is adaptive, i.e. it grows "smarter" as the sampling period increases so acceptable/normal activity range projections are more accurate for each point in time.
This time-sensitive baseline analysis:
is resource-light and runs in the middle tier
features configurable color alerts
supports full compare capabilities
features intelligent calibration of charts and bars display
The baseline perspectives are always relative to the latest norms, so if the instance changes, the baseline is re-applied. Thus, behavior which in its original context was reported as within a normative range can subsequently, and even retrospectively, be deemed exceptional.
Note: A performance baseline is not a definition of overall system performance; rather a realistic workload baseline gives valuable insight into expected real performance. This insight then allows database and administrative staff to effectively anticipate and plan for future needs.
Use case scenarios — identifying abnormal behavior
A key business problem solved by IntelliProfile is that of identifying, diagnosing and characterizing abnormal behavior. Almost all performance indicators are “relative”, but administrators must specify behavioral ranges specific to each of their instances. Are peaks presently encountered in fact exceptional? How exceptional are they? Do they reflect desired or undesired system responses? Are they one-time or cyclical? The baseline-generated exceptions trigger advisory alerts which help identify and correct abnormal instance behavior, and explain why exceptions occurred (what were the underlying causes).
See Tuning Scenarios for possibilities that lend themselves to baseline analysis.
Display and usage within Performance Analysis
The baseline graphic can be toggled on three different timeline views: on both the Home and History modes' Overview tabs in the instance workload view, and within the History mode's overall Resource time-line instance view.
Graphical interface:
The baseline range is displayed as a light
blue band in the background of the Home
mode > Overview tab's Instance Workload
graph and in the History
mode | Average Active Sessions timelines, the baseline range is displayed
as a background light blue band.
The baseline ranges are depicted as a range-marking
vertical element on the Overview Workload, Resource Breakdown and System
Utilization bar graphs..
The Overview, Workload, Resource Breakdown
and System Utilization bar graphs feature configurable
color alerts for behavior exceeding specified thresholds, which also
offer a hyperlink starting point for further analysis.