BUILD RECORD / 05
TIER 2 · FINTECH / ENGINEERING / DATA
BILLING DATA GATEWAY
Financial cost data should be normalized at memory speed, without a framework, without dependencies, and without losing a digit.
A zero-dependency C11 utility that normalizes multi-cloud cost exports (AWS CUR 2.0 and Azure) into an intermediate financial model, using POSIX memory address page projections rather than a heavier stack.
DIAGRAM ABOVE IS ILLUSTRATIVE — IT DEPICTS THE SYSTEM SHAPE, NOT A MEASUREMENT.
OVERVIEW
A zero-dependency C11 utility that normalizes multi-cloud cost exports (AWS CUR 2.0 and Azure) into an intermediate financial model, using POSIX memory address page projections rather than a heavier stack.
ZERO-DEPENDENCY FINANCIAL DATA ENGINE (C11)
PROBLEM
Multi-cloud cost exports arrive in different schemas — AWS CUR 2.0 and Azure Cost Management do not agree on shape, naming or granularity.
Reconciling them in a high-level runtime means paying a per-record interpretation cost on data volumes measured in millions of rows, and depending on a large toolchain to do it.
APPROACH
- Normalize both export formats into a single intermediate financial model at the point of ingestion.
- Eliminate the interpretation layer by projecting POSIX memory address pages directly over the input, so records are read in place rather than parsed into objects.
- Keep the build at zero third-party dependencies so the binary is the whole dependency graph.
ARCHITECTURE
ARCHITECTURE — DATA PATH
ARCHITECTURE DIAGRAM — THE DOCUMENTED LAYERS, NOT A MEASUREMENT
- 01
SOURCE EXPORTS
Untouched vendor cost exports
- AWS CUR 2.0
- Azure Cost Management
- 02
PAGE PROJECTION
Read in place, no copy
- POSIX mmap
- Address-page projection
- 03
FORMAT ADAPTERS
Per-source field mapping
- Column resolution
- Type coercion
- 04
INTERMEDIATE MODEL
Canonical financial record
- Single schema
- Query-ready
- 05
EMIT
Downstream consumable output
- Bulk write
- No framework
RESULTS
487,421
RECORDS / SEC
SELF-BENCHMARK — SELF-MEASURED THROUGHPUT
31.89
MB / SEC
SELF-BENCHMARK — SELF-MEASURED THROUGHPUT
2,051
NS PER RECORD
SELF-BENCHMARK — SELF-MEASURED LATENCY
The binary normalizes multi-cloud cost exports at self-measured throughput of 487,421 records/sec and 31.89 MB/sec.
Per-record cost measured at 2,051 ns on the project's own harness.
Dependency graph reduced to the C standard library and POSIX.
LIMITATIONS
- Throughput figures are self-measured on the project's own harness and are hardware, compiler and dataset dependent — they are not an independent third-party benchmark.
- Record and byte rates apply to the shapes benchmarked; other export variants or column mixes are not represented by these numbers.
- The C11 implementation trades operator ergonomics for speed; changing the record schema means recompiling, not reconfiguring.
TESTING
Correctness is exercised against representative export fixtures for both cloud providers.
Benchmark methodology and dataset characteristics are published alongside the figures in the repository.
ARTIFACTS
- Released v1.0.0 tag with the C11 source tree.
- Benchmark harness producing the throughput and latency figures quoted here.
- Normalization specification for the intermediate financial model.
CURRENT STATE
RELEASED v1.0.0
v1.0.0 released and available in its repository.
Figures come from the repository's published self-benchmark.