← All tools

Tool

Desktop variant for large journals

Measured with 100 million synthetic posting lines: about 27 minutes for a complete run.

Measured · Desktop

The desktop JET checks 100,000,000 journal lines.

That many rows do not fit into the memory of one computer. The desktop variant is built for exactly that. Measured with 100 million synthetic rows: about 27 minutes. The figures below are a model calculation.

The measurement with 100 million rows in detail →

Memory and runtime model

How memory and runtime change with the size of the journal. Move the sliders.

Raw data: –
1. Everything in memoryfits
–
grows with every row
On disk–
Runtime–
2. Spill to diskDisk
–
memory limit with a buffer
On disk–
Runtime–
3. Row by rowConstant
–
independent of journal size
On disk–
Runtime–

Peak memory compared

Logarithmic scale. The dashed line is the memory limit.

Memory limit
1. All in memory
2. Spill to disk
3. Row by row

Model calculation with 16 columns at 100 million rows as the anchor, scaled from the desktop measurement. Not a measurement on your computer.

Benchmarks · Desktop app

100 million postings on your computer

The app loads a journal, scores every posting against the audit criteria and delivers overview, drill-down and advanced analyses. Scale up your own volume and see where other tools stop.

100,000,000
Rows in one run, measured with synthetic data and a realistic distribution of accounts and users.
Complete in about 27 minutes.

How long does your journal take?

Enter the number of rows in millions or drag the slider. Times are scaled linearly from the measurement with 100 million rows.

Rows in the journal
M rows

27 min
for the full run: import, testing, advanced analyses and reports

    Memory: the database stays at 2 GB and puts the rest on disk. At 100 million rows we measured about 3 to 4.5 GB of process memory and several GB of temporary disk space.

    What happens

    Six steps, each with its own time. The times adjust to the amount you chose above.

    The account blocks split the journal by account. Each block can be checked on its own, so memory stays limited even for large journals.

    In comparison

    A spreadsheet and a browser evaluation each have a limit where the work tips over. The scale is logarithmic: each mark stands for ten times the previous amount.

    Spreadsheet
    Row limit of a workbook: 1,048,576 rows.
    Browser evaluation
    Smooth up to about one million rows, then work gets sluggish.
    Desktop app
    Solid: measured up to 100 million. Dashed: extrapolated.

    How it was measured

    All times come from one run with 100 million synthetic rows. Accounts, offsetting accounts and users are distributed realistically: 1,500 accounts, 1,500 offsetting accounts, 300 users. The database memory limit was 2 GB, 4 cores ran in a Linux container.

    StepSeconds (100 M)Seconds per M
    Import into the journal5395.39
    Special accounts and key figures60.06
    Splitting into account blocks1371.37
    Document totals450.45
    Document values per block770.77
    Check chain per block4544.54
    Testing total (JET)7277.27
    Advanced analyses total3613.61
    Overview, Benford, drill-down220.22
    • Advanced analyses: scoring and writing about 147 s, micro flags about 115 s, circle search about 63 s, learning sample and behaviour deviation about 35 s.
    • The extrapolation is linear. With smaller volumes the effort per row tends to be lower. The circle search and micro flags depend on the account structure.
    • On your computer the times may differ, depending on processor, disk and memory.

    The figures come from measurement runs on a Linux container with 4 cores. Status: November 2026.