Key Features
Innovative Design
Utilizes Write Ahead Log (WAL) for records, reducing IO and memory usage. Background compaction with Gorilla compression reclaims disk space.
Crazy Benchmark
Top performance. 19,172 ns/op. In-memory-like speed. WAL-class durability.
Super Easy Integration
Identical HTTP API and TCP interfaces, which are all in strict JSON.
Analytics Ready
Built-in support for data downsampling and aggregation.
Memory Efficient
As Low as 6MB memory. Perfect for IoT devices. Indexing is all in SSD.
Built-in Streaming
Subscribe to keys and receive updates in real-time.
Battle-Tested
Trusted by IoT pioneers and used in production. Code Coverage for all logic.
Cross-Platform
Supports Windows, Linux/BSD, and macOS. Perfect for edge devices.
Monitoring Ready
Built-in /health and /metrics (Prometheus) endpoints. Monitor uptime, memory, GC, and data points in real-time.
Batch Write
Write up to 10,000 data points in a single API call. Perfect for bulk imports and migration.
Data Export
Export sensor data in CSV or JSON format with filtering by time range and downsampling.
Gorilla Compression
Facebook Gorilla time-series compression. 8x smaller files, 56x faster writes during compaction.
Advanced Analytics
Downsampling with avg, sum, min, max, first, last, count, median (p50), p95, and p99 aggregations.
Usages
POST /
{
"operation": "write",
"key": "a_sensor1",
"write": {
"value": 32242424243333333333.3333
}
}
Need more details? Check out our complete API documentation.
View Full API DocumentationPerformance Comparison
Benchmark Results
Write Performance (ms)
Read Latest 100 Data 10 times (ms)
Read: 10k Queries (ms)
Multi-Write Performance (ms)
PubSub Performance (seconds)
Storage per 5,000 points (KB)
Test Configuration
| Total Data Points | 10,000 |
| Points per Sensor | 1,000 |
| Sensor Count | 10 |
| Write Method | Sequential, single-point |
| Read Method | 1 query (last 100 records) |
| Multi-Write Method | 10 goroutines parallel |
| PubSub message Count | 1,000,000 |
| Queue Order | Ordered |
| Queue Delivery Semantics | Exactly Once |
| OS | Windows |
| Architecture | amd64 |
| CPU | Core(TM) i7-13700KF |
Key Note
- GTSDB shows 233x faster write (21.76 ms vs 5,070 ms)
- Sub-millisecond single read (<1 ms vs 4.48 ms)
- 4.7x faster on 10,000 reads (205 ms vs 967 ms)
- 16.6x faster parallel multi-write (51 ms vs 851 ms)
- 29.6x smaller than JSON, 7.98x smaller than raw
- NSQ-like PubSub: 1M messages in 32.7s
- Only ~12MB Memory Usage
- Only 1 binary executable
GTSDB