vinclv / data-engineering-minds-kafka

This repository contains the components that I use for my Youtube Kafka videos
31 stars 41 forks source link

Single node cluster #1

Open BinSyed opened 3 years ago

BinSyed commented 3 years ago

@vinclv Thanks for your work.

I am working on a video analytics project where I have to detect 5 kinds of objects from the 'N' number of CCTV cameras. And, The customer provided only one PC ( i7 + 4TB SSD + GPU) system to deploy my Video analytics Engine.

As per your tutorials, I have set up Kafka + zookeeper + topic creation on my provided ubuntu PC.

This application going to be deployed on different site locations where the camera count ( N ) changes accordingly.

e.g) some locations they have only (N = ) 5 camera and other ( N= ) 25.

so per day average, my application has to handle 200GB / day data from cameras.

So I choose the below Big-Data tech stack to implement this,

image

in the above image ( HDFS Storage is not applicable, I am using Kafka as a Database as well Since we don't need any persistence storage at client side)

**Since I have only a single PC and single SSD drive,

  1. Is it OK to create multiple brokers on a single PC?
  2. Is it OK to create multiple partitions for a single topic?**

What happens if I use the " SINGLE BROKER + SINGLE PARTITION FOR SINGLE TOPIC + SINGLE CONSUMER " approach.

I am highly curious to know your advise on this,

Note: I know this will have a SINGLE POINT of FAILURE. but the customer agreed to proceed.