Performance Benchmarks & Technical Metrics
Core configuration parameters and design metrics, quantifying product capabilities
Core Configuration Parameters
System default configuration, adjustable per actual requirements
12
Thread pool max concurrency
Server version default
Configurable
1GB
Desktop / standalone version file limit
Subscription / Server version
Configurable
10GB
Network version file limit
Network version default
Configurable
100/60s
Default rate limit
Configurable
3-10
K8s HPA replica count
Supports auto-scaling
9
Docker Compose service count
Database · Cache · API · Monitoring · Frontend, etc.
2
Task queue count
Compression task · Maintenance task
15
Compression scenario count
Covers mainstream application scenarios
Design Metrics
Performance design goals for core security verification
<1ms
Authentication latency
Desktop / Mobile
<5ms
Authorization verification latency
Server-side
Measured Compression Ratios
Internal benchmarks on real-world samples, for evaluation reference
| Sample Scenario | Ratio | Level | Notes |
|---|---|---|---|
| ERP spreadsheet (xlsx with embedded objects) | up to 11.64x | Ultra | Recursive embedded-object compression + font subsetting + OLE bitmap compression |
| E-learning recording PPTX (embedded fonts ~40%) | 9.19x | Ultra | 57.6MB → 6.27MB via embedded font subsetting |
| OFD scanned document (447 pages, 67.7MB) | 2.92x | Ultra | 4~11x faster than the legacy engine, signature/attachment whitelist protection |
| Photo TIFF (uncompressed RGB) | up to 110x | Ultra | Native lossless channel, byte-identical pixels |
| Satellite GeoTIFF (palette LZW) | 1.03x | Ultra | Source already highly compressed — the lossless ceiling; georeferencing tags fully preserved |
| Text family (Markdown/YAML/GeoJSON etc.) | 20 ~ 54x | Balanced~Ultra | Tiered zstd/brotli lossless encoding, byte-for-byte restoration |
| Medical DICOM | 1.34 ~ 1.58x | Strong | Lossless re-sealing with the standard Deflated syntax, pixels untouched |
| Geospatial Shapefile bundle | 3.53x | Strong | shp/shx/dbf/prj bundled together, directly readable by GIS tools |
Ratio = original size ÷ compressed size. Figures are measured on internal real-world samples; actual gains depend on file content. All formats ship with a gain guard — when re-compression yields no benefit, the output is byte-for-byte identical to the input (never grows) — and lossy re-encoding passes SSIM≥0.90 / PSNR≥30dB quality floors.
Architecture & Deployment Parameters
System default configuration parameters
| Parameter | Default Value |
|---|---|
| Server version thread pool | 12 concurrent |
| Standalone version file limit | 1GB |
| Network version file limit | 10GB |
| Default rate limit | 100 times / 60 seconds |
| K8s HPA auto-scaling | 3 ~ 10 replicas |
| Docker Compose services | 9 |
Self-Testing Guide
Recommended steps to verify actual performance
-
Prepare test set: Select representative file samples (e.g., 10 photos of different resolutions, 5 PDFs of different sizes, 3 videos of different durations) covering your main usage formats.
-
Record original size: Record the original byte count of each file as the basis for compression ratio calculation.
-
Execute compression: Use the desktop version or call the SDK/API to perform compression, select the target compression level and scenario, and record compression time (using the
timecommand or in-code timing). -
Calculate compression ratio: Compression ratio = (original size - compressed size) / original size × 100%. It is recommended to aggregate statistics by format rather than mixing them together.
-
Quality assessment: For images/videos, compare PSNR or SSIM metrics to assess quality loss; for documents, manually check whether key content is clearly readable.
-
Concurrent testing (Server version): Use stress testing tools to send concurrent requests to the API and observe thread pool utilization and response latency.
Need a customized performance evaluation?
Contact us to test against your actual file scenarios