Validated Performance Data

Real hardware. Real workloads. All results independently reproducible.

AriaOS — Governance Layer

Platform: Jetson AGX Orin 64GB, Apple Silicon M-series, HP ProLiant G8 | Version: v2.5.3 Stable | TRL 6 Validated

132.6/100 Composite benchmark score (Jetson AGX Orin)

47ms P95 response latency

99.97% Uptime — 800/800 endpoints validated

2,847 RPS Requests per second capacity

Memory Bus Performance

Internal message bus benchmarks — AriaOS governance layer

Metric	Target	Measured	Status
Message Latency (P50)	<1ms	0.4ms	PASS
Message Latency (P95)	<5ms	2.1ms	PASS
Throughput	>10K msg/s	47K msg/s	PASS
Memory Overhead	<64MB	42MB	PASS

Autonomous Recovery

Metric	Value
Detection Time	<100ms
Isolation Time	<500ms
Full Recovery	<2 seconds
Data Loss Events	0

Validation Environments

Jetson AGX Orin 64GB

132/100 composite score. Hardware validation and inference benchmarks on NVIDIA edge AI platform.

Apple Silicon (14+ days)

Sustained fault injection, concurrency, governance enforcement, and audit integrity on M-series SoC.

HP ProLiant G8 Chaos

Ubuntu 24.04 — network degradation, resource exhaustion, recovery behavior, governance continuity.

Source: AriaOS.dev/docs | Memory stability: 99.8% | 0 cloud dependencies

HammerIO — Compression Engine

Platform: Jetson AGX Orin 64GB | JetPack 6.2.2 | CUDA 12.6 | MAXN mode | All results roundtrip verified

8,537 MB/s GPU decompress (in-memory)

4,258 MB/s GPU decompress (10GB roundtrip)

5.8x Roundtrip speedup vs CPU zstd-1

In-Memory Performance (Raw Throughput)

Method	Processor	Compress	Decompress	Integrity
nvCOMP LZ4	GPU	705 MB/s	8,537 MB/s	PASS
nvCOMP Snappy	GPU	1,615 MB/s	5,756 MB/s	PASS
zstd-1	CPU	1,747 MB/s	2,001 MB/s	PASS

Roundtrip Results (10GB with Disk I/O)

Method	Processor	Compress	Decompress	Ratio	Integrity
nvCOMP LZ4	GPU	517 MB/s	4,258 MB/s	1.98x	PASS
zstd-1	CPU	1,094 MB/s	733 MB/s	2.00x	PASS
zstd-3	CPU	1,014 MB/s	741 MB/s	2.00x	PASS

ModelSafe — Model Checkpoint Manager

Platform: Jetson AGX Orin 64GB | Engine: nvCOMP GPU LZ4 via HammerIO | 2026-04-06

3.6s 7B model restore time

7.9s 13B model restore time

391 MB/s Peak decompress throughput

Model Size	Original	Compressed	Ratio	Compress	Decompress	Restore Time	Integrity
1GB (7B equiv)	1.0 GB	875.1 MB	1.17x	208 MB/s	283 MB/s	3.615s	PASS
3GB (13B equiv)	3.0 GB	2.4 GB	1.24x	237 MB/s	391 MB/s	7.861s	PASS
7GB (70B equiv)	7.0 GB	5.7 GB	1.23x	289 MB/s	378 MB/s	18.979s	PASS

Note: Synthetic model weight data — float32 arrays with realistic distribution. Ratios reflect actual model weight compressibility. Restore time includes SHA-256 verification. Powered by HammerIO nvCOMP GPU LZ4.

AriaOS: Forge — Domain Model Training

Platform: Jetson AGX Orin 64GB | Base: Qwen2.5-Coder-3B | Method: LoRA fp16 | Dataset: OpenStack Q&A | 2026-04-06

~10h Full pipeline — end to end, no cloud

19.7 tok/s Fine-tuned model inference speed

33% Quality improvement over base 7B model

A/B Comparison — Fine-tuned vs Base Model

Model	Parameters	Quality Score	Tokens/sec	TTFT	Result
ariaos-forge:latest	3B (LoRA fine-tuned)	80/100	19.7	4,467ms	WINNER
qwen2.5-coder:7b	7B (base, untuned)	60/100	10.3	29,853ms	—

Note: Quality scoring via PraetorianMind Model A/B Compare — heuristic evaluation only. Speed advantage reflects 3B vs 7B parameter count. Training dataset: 623 samples (OpenStack Q&A + documentation). All training performed locally — zero cloud dependency.

Validation Environments

Platform	OS	CUDA	Mode	Modules Tested
Jetson AGX Orin 64 GB	JetPack 6.2.2	12.6	MAXN	All
Apple Silicon M-series	macOS	N/A	Max	AriaOS
HP ProLiant G8	Ubuntu 24.04	N/A	—	AriaOS chaos

Explore the platform →