Recognition Server

Attention

For stable operation of the server, databases, and system components, your system should meet all of the listed requirements. Failure to meet any of the requirements may cause the system to work incorrectly.

System requirements for the graphic recognition server:

Recommended System Requirements

Description

Operating System

  • Ubuntu 18.04;

  • Ubuntu 20.04;

  • Ubuntu 22.04;

  • Astra Linux 1.7.

Additional Software

  • Python 3.7;

  • Python 3.7-venv;

  • libpoppler-cpp0v5 poppler-utils libsm6 tesseract-ocr.

    Check Python version
    Check Python version: python3 --version

    Install Python3.7: sudo apt install python3.7

CPU

Intel or AMD with AVX support. The number of threads/kernels depends on the number of agents.
One kernel can process up to 28,800 images per day.
On average, an agent generates about 2000 images per day.

Disk Size

up to 100 GB.

Disk

NVMe with a read/write speed of at least 1500 MB/sec.

IOPS

at least 5000 I/O operations per second.

RAM

Depends on file size and complexity.
It is recommended to allocate at least 80 MB per image or screenshot,
up to 300 MB.

CPU and RAM requirements for daily data processing, including large files:

Agent Amount

Average Image Amount

CPU

RAM

10

20 000

2

4 GB

50

100 000

6

6 GB

100

200 000

9

6 GB

200

400 000

16

8 GB

500

1 000 000

37

16 GB

1000

2 000 000

72

24 GB

Calculating Threads/Kernels and RAM

Kernel/threads calculation:

Note

If the Only make screenshot when process changed option is turned off, turning on the Screenshots on changing window focus option will generate a very large amount of screenshots.

A user typically generates 1,500 to 2,000 screenshots per day, depending on their activity and the number of monitors. One kernel can process up to 28,800 screenshots per day.

The example below assumes 300 PCs with the Screenshots module turned on.

In that case, the amount of screenshot per day would be:

2000 * 300 = 600 000 screenshots

The total number of screenshots divided by the number of images processed by one core:

600 000 / 28 800 ≈ 21 threads/kernels

RAM Calculation

On average, image processing requires about 80 MB per kernel. To calculate the amount of RAM, multiply the number of kernels/threads by the memory requirements for one kernel and one thread:

21*80 = 1680 MB

The image converters and the server operating system work in parallel. The converters consume from 1 to 4 GB of RAM, depending on the data amount.

For optimal system operation under heavy load, use a server with at least 8 GB of RAM.