View on GitHub

Overlord

The Python and Vagrant part of TSDBBench

Uses LevelDB for indexes and uses a custom storage layer when persisiting data on the local filesystem
“multi-dimensional data model”
Features a flexible query language
- Recording rules to save results of frequently needed expressions in separate time series
Indirect pushing of time series data via pushgateway
- Prometheus-Clients available to push data of the push metrics types counter, gauge, histogram or summary
  - Supported exposition format is the Protocol buffer format
  - No support of own timestamps, time of push is used instead
- HTTP API to push to pushgateway
  - Supported exposition formats are plain text format and Protocol buffer format
  - Own timestamps can be used when using plain text format
Two different query formats of ranges
- Using offsets from the currect point of time and a granularity of s, m, h, d, w, y from there to include the desired time series data in a vector
  - Can be used by buitlin functions to calculate values as sum, avg, min, max
- Using two timestamps rfc3339- or Unix-timestamps and a step size of ms, s, h, d, w, y to to include the desired time series in a matrix
  - No further postprocessing possible with builtin functions
Prebuilt releases for prometheus and pushgateway are available

As custom timestamps are used in insert steps, the plaintext format was used
As postprocessing functions are used in the scan steps, the vector result format was used
- Timestamps were used to calculate offset and granularity to include the desired range of time series and overestimate the range as little as possible