Now on Google Cloud Marketplace — start a trial in your own project
Research & publications

We put it in the record.

DataForge's claims are not marketing — they are documented, falsifiable, and peer-reviewable, published under a permanent DOI with the methodology and dataset attached.

Whitepaper · Zenodo · open access (CC BY 4.0)

Pressure Curves and Bottleneck Migration: Hardware Limits Revealed By The DataForge

Author · Osei Harper Published · 2026 DOI · 10.5281/zenodo.19503175 ORCID · 0009-0004-5771-0406

What the paper shows

  • 01The engine is never the bottleneck. Across local, enterprise, and cloud environments, throughput is bounded by the destination protocol, the storage write ceiling, or CPU-allocation policy — not by DataForge.
  • 02Bottleneck migration. On a Pure Storage FlashArray X90R4 the limit moved off storage entirely and onto CPU, where the engine still reached 8,151,597 rows/sec sustained.
  • 03A NUMA anomaly in SQL ingestion under high concurrency — surfaced and documented empirically, with the per-concurrency pressure curves.
  • 04Deterministic WAL output across heterogeneous environments — the same run reconciles to the same result, machine to machine.

Reproducible by design

The study is built on a real public dataset — the Free Law Project's CourtListener corpus — run under observable, repeatable conditions on disclosed hardware. The benchmark harness is freely downloadable, so the central results can be reproduced on a comparable hardware class by anyone who cares to check. That is what "peer-reviewable" means here: not that a journal has stamped it, but that the evidence is open to being checked — and structured so that it can be.

How to cite

Harper, O. (2026). Pressure Curves and Bottleneck Migration: Hardware Limits Revealed By The DataForge. Zenodo. https://doi.org/10.5281/zenodo.19503175
Scholarship

An independent-scholar practice

The paper is one output of a broader independent-scholar corpus spanning systems theory, epistemics, and the philosophy of technology — the same discipline that produced credited entries in the U.S. National Vulnerability Database (CVE-2025-3462, CVE-2025-3463). DataForge is engineering held to a research standard: state the claim, publish the method, invite the check.

Contact us