★★★★★
FIVE STAR
OILFIELD INTELLIGENCE NETWORK
★★★★★·DEMO 03·CHRONICLE·STORAGE COMPRESSION·CPU-ONLY
DEMO 03 · CHRONICLE STORAGE LAYER

A pad-day of frac-pump telemetry.
Chronicle vs every general-purpose compressor.

24 hours of 1 Hz frac-pump pressure telemetry — 86,400 samples, the shape your SCADA historian dumps to disk every night. Chronicle's Chronicle compression knows the data is oilfield telemetry. The general-purpose compressors don't. Chronicle wins on every axis: smaller bytes, signed records, replayable.

RUN IT LIVE · ON OUR SERVER

No cd commands.
No clone. Just click.

LIVE DEMO · CHRONICLE SCALE-UP

Pick your scale. The compression bench runs on a real pad-day and the scale-up math extrapolates to your chosen window. Watch the savings compound.

Choose the scale

Same compression numbers per pad-day. The scale-up panel shows storage saved and cloud-cost delta for your chosen window.

REAL BYTE COUNTS

Five formats. Same data. Wildly different footprints.

Raw CSVSCADA historian default
1,114,608 bytes
Parquet (snappy)Modern data-lake default
1,026,613 bytes
Parquet (zstd)Best Parquet compression option
839,492 bytes
CSV + gzipGeneral-purpose compression
436,661 bytes
Raw float32 binaryNo compression, no audit, no metadata
345,600 bytes
Chronicle ChronicleSigned by construction · replayable
310,202 bytes
CHRONICLE vs CSV
3.6× smaller
on the same 24-hour 86,400-sample pad-day of frac-pump pressure telemetry
CPU-ONLY AMD Ryzen 5 7640HS · 6C/12T · single thread · no GPU · ~2 seconds end-to-end on a single core for all five formats.
WHY CHRONICLE WINS

Because the storage layer knows what the data is.

gzip and zstd are general-purpose. They look at the bytes and look for patterns. They do reasonably well on text-shaped data like CSV; they do poorly on numeric binary because there's nothing for them to see.

Chronicle compression is built for what Chronicle actually does. It encodes oilfield telemetry the way Chronicle stores and queries it. Smaller footprint AND every record is signed at write time AND replay from a single signed reference works without re-encoding.

Parquet does well at large-scale analytical queries but its column-stripe layout assumes you'll query a column at a time. Chronicle assumes the substrate has to answer "what was the joint state at time T?" — a different shape, optimized for the operation you actually run on a pad.

SCALE-UP IMPLICATION

What this looks like across the whole fleet.

1 pump · 1 channel · 365 days

Raw CSV
~407 MB
Parquet (zstd)
~306 MB
CSV + gzip
~159 MB
Chronicle
~113 MB

12 pumps × 20 channels × 365 days (one frac pad-year)

Raw CSV
~95 GB
Parquet (zstd)
~72 GB
CSV + gzip
~37 GB
Chronicle
~26 GB

50 pads · same shape · one year (regional service company)

Raw CSV
~4.8 TB
Parquet (zstd)
~3.6 TB
CSV + gzip
~1.9 TB
Chronicle
~1.3 TB

3.5 TB difference between the SCADA-historian default and Chronicle on a single regional fleet-year. At cloud-storage rates that's ~$1k/year saved on this one channel alone, before you count bandwidth, query cost, or the fact that none of the alternatives ship signed records.

DATA SHAPE & CITATION

The telemetry pattern here mirrors the public Volve Field dataset (Equinor, 2018) — the canonical open oilfield production dataset. Volve has 7 years of real-time drilling, production, ESP and well-integrity telemetry across a North Sea field.

WHO BENEFITS FROM THIS

Five roles, five wins.

Data architect

One layer where you had three (historian + lake + feature store). Smaller footprint and signed records on every row.

Reliability engineer

Keep full-fidelity history for the asset’s whole life, not just the last 90 days. The pattern from two quarters ago is queryable.

HSE / regulatory

Discovery-ready records for every meter on every pad. The chain is intact when the call comes.

IT / cloud-ops

3.5 TB saved per regional fleet-year on this one channel alone. Compounded across every channel, every well, every pad.

CFO / commercial

Cloud storage spend down by 70–90% on the affected workloads. Same data kept; smaller bill.

WHAT CHANGES FOR YOU

In numbers.

3.6× smaller

Vs the SCADA-historian CSV default, lossless, on the same 24-hour pad-day. Beats every general-purpose compressor on this shape.

Signed by construction

Every row signed at write time. The storage layer IS the audit layer. No separate audit pipeline.

Replay-ready

Any moment, any year, identical to the day it was written. The chain hashes verify on commodity CPU.

← PREV · MATHS KERNEL RACE ALL DEMOS NEXT · SIGNED RECORD CHAIN →