tlas.atlas-wiki.1.4.4.source-code.Concepts.md Maven / Gradle / Ivy
The newest version!
### Aggregation Function
Combines a set of time series into a single time series.
A conceptual measurement such as cpu usage or requests per second. A metric will usually end up getting reported as a set of time series, for example a cpu usage time series for each node.
### Consolidation Function
Converts a time series from a primary step size to a larger step size that is a multiple of the primary. For example, if you have a time series with data points each minute and you want a time series with data points every 5 minutes.
In most cases you won't use a consolidation function directly. The system will automatically apply consolidation to ensure there is at least one pixel per datapoint on the rendered images.
### NaN Aware Operations
This typically refers to how binary operations treat NaN values in the data. A NaN value means that the value is unknown for a given interval. NaN aware operations will keep the value as NaN if both sides are NaN, but treat it as a special value such as 0.0 otherwise. This is useful to avoid gaps in the graph if a subset of time series used in an aggregate have gaps in data.
| *lhs* | *rhs* | *result* |
|-------|-------|----------|
| NaN | NaN | op(NaN, NaN) |
| v1 | NaN | op(v1, 0.0) |
| NaN | v2 | op(0.0, v2) |
| v1 | v2 | op(v1, v2) |
### Naming Conventions
See [naming conventions](https://github.com/Netflix/spectator/wiki/Conventions).
### Normalization
In Atlas this usually refers to normalizing data points to step boundaries.
### Time Series
Metadata plus a set of data points reported at a regular interval. In Atlas the metadata is a set of string key value pairs called tags. A data point is a pair consisting of a timestamp and value. For Atlas the value is typically a double.