-
Notifications
You must be signed in to change notification settings - Fork 3
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
sid_terrafloww
committed
Jan 5, 2025
1 parent
84f913e
commit aa0bc64
Showing
2 changed files
with
64 additions
and
1 deletion.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,63 @@ | ||
# Performance Comparison | ||
|
||
## Overview | ||
|
||
Rasteret has been benchmarked against traditional rasterio-based approaches, showing significant performance improvements for Sentinel-2 data access and processing. | ||
|
||
## Benchmark Results | ||
|
||
### Time Series Analysis Test Case | ||
Processing a year of Sentinel-2 data (20 scenes) for NDVI calculation: | ||
|
||
| Implementation | Processing Time | Memory Usage | | ||
|----------------|----------------|--------------| | ||
| Rasteret | ~8 seconds | Optimized | | ||
| Rasterio Direct | ~24 seconds | Higher | | ||
|
||
### Key Differences | ||
|
||
1. **Data Access Strategy** | ||
- Rasteret: Optimized tile-based access with parallel processing | ||
- Rasterio: Full window reads with masking | ||
|
||
2. **Memory Management** | ||
- Rasteret: Efficient tile caching and memory usage | ||
- Rasterio: Larger memory footprint due to full window reads | ||
|
||
3. **Processing Pipeline** | ||
- Rasteret: Integrated STAC metadata and optimized band access | ||
- Rasterio: Direct band access and processing | ||
|
||
## Implementation Comparison | ||
|
||
Both implementations produce identical NDVI values, validating the accuracy of Rasteret's approach while delivering superior performance. | ||
|
||
## Detailed Performance Analysis | ||
|
||
### Time Series Analysis (20 scenes) | ||
|
||
#### Rasteret | ||
- STAC filtering: ~0.5s | ||
- Data processing: ~7.4s | ||
- Total time: ~8.0s | ||
- Average per scene: ~0.4s | ||
- Memory usage: Optimized (~200MB) | ||
|
||
#### Traditional Rasterio | ||
- STAC search: ~2s | ||
- Data processing: ~23.1s | ||
- Total time: ~24.6s (with GDAL configs set for ideal performance) | ||
- Total time: ~44.6s (without GDAL configs set) | ||
- Average per scene: ~1.2s to ~2.2s | ||
- Memory usage: Higher (~500MB) | ||
|
||
### Performance Breakdown | ||
|
||
1. **Metadata Phase** | ||
- Rasteret's one-time metadata creation enables faster subsequent queries | ||
- STAC filtering is 3x faster with pre-cached metadata | ||
|
||
2. **Data Processing** | ||
- Rasteret: ~0.4s per scene | ||
- Rasterio: ~1.2s per scene | ||
- 3x improvement in per-scene processing |