blob: 2dd6df6cfb6379f765a1818fc3d07c6d4b7aae19 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
|
<!--
title: "Netdata storage number"
custom_edit_url: https://github.com/netdata/netdata/edit/master/libnetdata/storage_number/README.md
sidebar_label: "Storage number"
learn_status: "Published"
learn_topic_type: "Tasks"
learn_rel_path: "Developers/libnetdata"
-->
# Netdata storage number
Although `netdata` does all its calculations using `long double`, it stores all values using
a **custom-made 32-bit number**.
This custom-made number can store in 29 bits values from `-167772150000000.0` to `167772150000000.0`
with a precision of 0.00001 (yes, it's a floating point number, meaning that higher integer values
have less decimal precision) and 3 bits for flags.
This provides an extremely optimized memory footprint with just 0.0001% max accuracy loss.
|
