Downsampling

Downsampling reduces the number of data points returned in query results while preserving visual fidelity. It is applied as a post-processing step in the Router after query results are merged.

Algorithms

Mode	Algorithm	Description	Points per Bucket
none	—	No downsampling	All
auto	Auto-detect	Selects best algorithm based on data spikiness	Varies
lttb	Largest-Triangle-Three-Buckets	Best visual fidelity for smooth data	1
minmax	Min-Max per bucket	Preserves peaks and valleys	2
avg	Average per bucket	Fast, computed averaged values	1
m4	First-Min-Max-Last	Best shape preservation	4

Auto Mode

When downsampling=auto, the system analyzes data characteristics to select the best algorithm:

1. Calculates a spikiness score (0–1) based on:

   • Absolute deviation > 2σ from mean
   • Derivative spikes (rate of change > 1σ)

2. Selects algorithm based on data size and spikiness:

Condition	Algorithm
>100K points + spiky (>0.2)	MinMax
>100K points + smooth	Average
Spikiness > 0.2	MinMax
Spikiness > 0.1	M4
Smooth data	LTTB

Default auto threshold: 1,000 points.

Usage in Queries

Downsampling is configured via query parameters:

# Auto downsampling to 500 points
curl "http://localhost:5555/v1/.../query?\
device_id=sensor-001&\
start_time=2026-01-01T00:00:00Z&\
end_time=2026-02-01T00:00:00Z&\
downsampling=auto&\
downsampling_threshold=500" \
  -H "X-API-Key: your-api-key"

# Explicit LTTB downsampling
curl "http://localhost:5555/v1/.../query?\
downsampling=lttb&\
downsampling_threshold=1000" \
  -H "X-API-Key: your-api-key"

Downsampling vs Aggregation

These are different concepts in Soltix:

Feature	Aggregation	Downsampling
When	Pre-computed on flush	Applied at query time
Where	Storage service	Router post-processing
Purpose	Speed up time-range queries	Reduce response size for visualization
Levels	Fixed: 1h, 1d, 1M, 1y	Configurable threshold
Output	sum, avg, min, max, count	Original or reduced data points