AI Ingestion

Gemini Copilot

Performance Baseline

The Parallel Ledger is engineered for consistent, high-density witnessing, even on standard commodity hardware. The following benchmarks represent a "Hashed Commit" load where every attribute is cryptographically processed, deduplicated, stored, and witnessed.

The Reference Environment

Benchmarks were conducted on consumer-grade silicon to establish a "conservative floor" for performance expectations.

Hardware Specs

  • CPU: AMD Ryzen 5 7600X (6C/12T)
  • RAM: 32GB DDR5
  • Storage: Samsung 990 NVMe (Gen 5)

Software Stack

  • OS: Debian VM (Proxmox)
  • Database: Postgres 17
  • Runtime: Erlang/OTP 27+ (BEAM)

Sustained Ingestion Performance

These figures reflect a continuous 1,000,000-update burst. This test utilizes batched ingestion (50 updates/commit) across 200 concurrent worker lanes to measure the engine's absolute I/O ceiling.

Metric Measured Value
Total Witnessed Updates 1,000,000
Unique Dictionary Population 2,001 Slugs / 2,000 Values
Cryptographic Throughput 1,000,950 Unique Hashes
Witnessed Block Density ~2,050 Facts / Block
Sustained Throughput ~10,734 Hashed Commits / Sec

Architectural Invariants

  • 01
    Linear Vertical Scaling Utilizing the BEAM’s native concurrency, throughput scales linearly with core count. On enterprise AMD EPYC or Intel Xeon platforms, the ceiling expands to saturate the storage controller before reaching CPU bottlenecks.
  • 02
    Mechanical Sympathy The engine's logic is decoupled from the I/O path. By avoiding expensive Tricode collisions during ingestion, the Ledger maintains predictable latency even during massive spikes in institutional data volume.
  • 03
    Density Optimization String deduplication and dictionary mapping ensure that 1,000,000 updates do not lead to 1,000,000 rows of storage bloat, preserving high performance over decade-long retention periods.