Community
- Command Central
- MVP Program
- Monthly Mission
- Blogs
- Groups
- Events
- Media Vault
Products
- Observability
- Network Management
- Application Management
- IT Security
- IT Service Management
- System Management
- Database Management
Content Exchange
- SolarWinds Platform
- Server & Application Monitor
- Database Performance Analyzer
- Server Configuration Monitor
- Network Performance Monitor
- Network Configuration Manager
- SQL Sentry
- Web Help Desk
Free Tools & Trials

Help with Perfstack Display Transforms

Can someone point me in the direction of where I can find some documentation or guidance on the Display Transforms that are available in Perfstack Analysis? I would love to see a set of examples and how to effectively utilize these. (Change Point, Difference, Linear Regression, Normalized, Percentile Standardized, Smoothing, Smoothing/Standardized, Standardized.)

I do not understand what they do and why they change the values. It has been too long since I cracked a statistics textbook!

Find more posts tagged with

PerfStack

Accepted answers

silverbacksays

Hey dilbert1234

Check out this section in the NPM Administrators Guide:

Troubleshooting environmental issues with Performance Analysis dashboards

It explains all about the various transforms, and how they can help you visualise your data.

All comments

silverbacksays

Hey dilbert1234

Check out this section in the NPM Administrators Guide:

Troubleshooting environmental issues with Performance Analysis dashboards

It explains all about the various transforms, and how they can help you visualise your data.

tomiannelli

dilbert1234 here is a copy of that section silverbacksays linked above.

Display Transform algorithms

You can apply any one of the following algorithms to a charted metric:

Change Point - useful for finding significant changes in the numerical average of a set of data over time. When the numerical average for a time series changes significantly, the charted line will move up or down to the new numerical average, indicating that there has been a notable statistical change in the numerical average of the data.

Difference - Shows the relative positive or negative difference between a data point and the point immediately following it. The largest differences indicate significant changes in data during the scope of the charted timeframe. If the metric being transformed is a percentage, then the absolute difference (which is always positive) is shown.

Linear Regression - Shows the overall linear trend of the charted data. The slope of the linear trend indicates the overall trend of the data in the charted timeframe. Linear Regression is useful for showing if a noisy set of data is trending up, trending down, or remaining constant.

Normalized - Compresses a charted timeframe by representing each data point as a value between 0 and 100, relative to each other data point in the set. 100 represents the highest data point, and 0 represents the lowest data point. Normalized data is useful for comparing metrics of different scales: since each normalized metric shows all data in a common scale of 0 to 100, the data becomes easier to compare.

Percentile Standardized - takes the data and calculates the median value and the median absolute deviation of the data set and calculates where each data point is relative to the median and median absolute deviation. The resulting range of data is indicative of anomalies from a percentile stand point. This can be useful because a mean value is often skewed by extreme outliers whereas percentiles are resilient to outliers in a large enough data set. If the metric being standardized is already a percentage, then the data will also be normalized in addition to being standardized.

Smoothing - Takes a moving average of the data so that odd spikes in the data are averaged out, leaving only the overall pattern of data. Smoothing is useful for finding patterns in data that has a lot of spikes.

Smoothing and Standardized - Combines the smoothing and standardized algorithims. Useful for finding cause and effect patterns. If the metric being standardized is a percentage, then the data will also be normalized in addition to being standardized.

Standardized - Useful for quantifying how anomalous a value is relative to the rest of the data in the charted timeframe. Viewing multiple timeframes of data in standardized format can help you see patterns where anomalous values or trends happen. Values between -3 and 3 are expected whereas values less than -3 or greater than 3 can be considered anomalous. If the metric being standardized is a percentage, then the data will also be normalized in addition to being standardized.

serena

did you end up ultimately using these? How did it go?