Measured, not promised
The speed is in the design, not in the hardware.
The usual ways to make a load faster are a bigger machine or fewer checks. Nodwise needs neither: every node checked against a closed grammar, on the base configuration our service runs on.
23.6 min
28 min
3%
Don’t take our word for it
How long does your warehouse’s lineage take to load in the tool you use today?
We are not going to tell you we are faster. Time it. Load the lineage you already have into your current tool, with the checks switched on, and put the number next to ours. Your data, your format, your tool — we publish the shape of our test so you can compare like with like.
Your run
Your tool
—
Nodwise
11.2
4 of 4 checks
Type your numbers. Compare at a similar size — the cost per node grows as any store grows — and with the same checks on both sides.
The shape of the test
A time means nothing without the shape it was measured on.
The test is a data catalog: schemas that contain tables, tables that contain columns, and lineage from each column to a column in the next schema. A richer shape — more properties per node, relations that carry properties, responsibilities — costs more. Compare against this shape, not against the headline.
Node type
Nodes
Property values
Relations
Schema
128
128
16 384
Table
16 384
16 384
2 097 152
Column
2 097 152
6 291 456
2 080 768
Total
2 113 664
6 307 968
4 194 304
Checked
Transactional
Recorded
Both ends
The run, step by step
Where the time goes — and where it does not.
The steps run in this order for every file and do not overlap, so they add up to the total. A first load is spent writing. A re-read is spent reading what is stored, only to find it unchanged. Checking against the grammar is never the expensive part.
#
Step
First load, seconds
Re-read, seconds
1
2
3
4
5
6
7
8
9
10
+
Total, in seconds
1 416
1 692
Your turn
Time your own load. Then see what an agent does with ours.
A fully modelled organisation, one click with your work account. No form, no call.