Minimum System Requirement

[NistantriTech-Dhaval]

What will be minimum system requirement for thingsboard pe docker with default setting in this repo?
we have aws m6a.xlarge which has 16GB ram with 4 vCpu and it shows 83% ram usage without much load
is it normal normal usage ?

[smatvienko-tb]

Hi!
The memory consumption depends on which setup you choose (basic, advanced). By default, the memory management is not set. And it will consume all the memory you have without boundaries set. Most of the micro services is a Java application. The best practice is to set the exact heap memory and allocate it at the startup. Please, set your own memory size on each microservice depends on your needs.

# .env
# Limit memory usage for each Java application
# JAVA_OPTS=-Xmx2048M -Xms2048M -Xss384k -XX:+AlwaysPreTouch

For JavaScript based (node.js) micro services like js-executor, web-ui, web-report you can use env like

NODE_OPTIONS="--max-old-space-size=200"

[t0mZ4]

hi @smatvienko-tb so for a 16GB machine with 4CPU (like amazon m6i.xlarge), what would you recommend to setup for BASIC mode ? the cluster is handling about 2000devices at 6msgs per sec

it is working fine but i want to keep 1-2gb of ram free

[smatvienko-tb]

Hi, @t0mZ4 . There is no magic config for 2k devices. For your load you are likely found your system mostly idling. But maybe your use cases do some heavy computations, aggregations, db queries, api calls, etc. So the resource consumption is quite individual.

To validate and update your instance according to you need - you can check your instance CPU/Memory metrics, Volume(s) metrics, DB metrics (AWS console or Grafana)

To optimize ThingsBoard memory usage - deploy Grafana dashboard like JVM (Micrometer)and check Java memory utilization

In general, having a single 4/16 memory m7i.xlarge instance is quite comfortable to maintain (not labour intensive) . with monthly on-demand cost ~150$/mo.
What options do you have to reduce costs?

• r7i.large (2/16) ~ $100/mo -- slower startup, but still have plenty of memory
• c7i.xlarge (4/8) ~ $130/mo -- half of memory, will hard to deploy monitoring, tools and other
• m7i.large (2/8) ~ 75$/mo -- slower start, will hard to deploy monitoring, tools and other

Other option to save costs is to get some Saving Plan (like 1year, no-upfront) option and get a discount for a long contract.
From my experience with AWS the EC2 is not the biggest expense in the bill.

The instance you provided is not the latest generation. If you choose the older instance because it cheaper - you better take a look on AMD m6a.xlarge (easy replcement) or ARM architecture (need to reinstall OS) m6g.xlarge

Anyway it is up to you how to solve the equation labor*x + hardware*y = profit.

Wish you an effortless journey with ThingsBoard!