Having thousands of devices slows NetAlertX functions considerably

[SCUR0]

Is there an existing issue for this?

• I have searched the existing open and closed issues

Is your feature request related to a problem? Please describe

I have over 9k devices if tracking guest VLANs.
Clicking "Devices" takes 12+ seconds.
Typing in the filter takes minutes to show. Partly because it searches while you type every character.

Describe the solution you'd like

Index (dashboard) should show device count summary per subnet. This would help reduce the slowdown just reaching the home page of NetAlertX.
Click a subnet to view devices in that subnet.
Search should wait for you to finish typing or only run after pressing enter or a search button.
Clicking "Network" is off the table. I don't even know what that feature looks like on my network. It never loads.

Describe alternatives you've considered

I increased RAM and CPU on the VM that the docker container is on. No noticeable difference.

Anything else?

I work for a school district and was looking for alternatives for Lansweeper. NetAlertX looks like a great replacement but it wasn't designed with thousands of devices in mind.
I'm hoping that an optimization pass for this isn't too much of a stretch for this project.

Am I willing to test this? 🧪

• I will do my best to test this feature on the netlertx-dev image when requested within 48h and report bugs to help deliver a great user experience for everyone and not to break existing installations.

Can I help implement this? 👩‍💻👨‍💻

• Yes
• No

[jokob-sk]

Thanks for the feedback.

I'm pretty impressed it's still usable with 9000 devices.

It could try moving the logs into tmpfs

      - type: tmpfs
        target: /app/log

If you don't care about historical data you could also move the DB there. The app/config folder contains the settings and devices backup.

      - type: tmpfs
        target: /app/db

Long term I am thinking supporting more perfomant DB engines.

Additional reading:

• https://jokob-sk.github.io/NetAlertX/PERFORMANCE/?h=performance
• https://jokob-sk.github.io/NetAlertX/BACKUPS/?h=backu#backup-strategies

Some notes:

Typing in the filter takes minutes to show. Partly because it searches while you type every character.

The search triggers after 3 characters and 0.5 s inactivity from memory. I can increase the delay if that helps.

[SCUR0]

It could try moving the logs into tmpfs

Is this the cause of some of the slowdown? I wasn't aware of any issues with the logs. Once I get everything stable I usually turn the logs to "none" in the core settings.

The search triggers after 3 characters and 0.5 s inactivity from memory. I can increase the delay if that helps.

Not sure if the slowdowns are causing issues with the inactivity timer but with a quick test, I typed 10.227 with no interrupts.
For a brief moment I saw results of searching 10.22 before it showed results for 10.227. 0.5 might be too short.

I was able to get devices down to 4k by removing some guest and student VLANs that don't need to be tracked.
It is handling that amount at a more respectable rate.

Are there any plans to make any kind of results/summary page based on subnets rather than having them all listed at once?
I'm unsure if that is what "Network" does, as it never loads for me. I assumed because it is trying to render thousands of devices.

[jokob-sk]

Thanks for the suggestions @SCUR0 .

I'll look into improving the performance in the future. For now I increased the search debounce delay to 750 ms (hope that won't bother others :) ) - hopefully this is a bit more manageable.

There might be a bigger front end rewrite coming so avoiding doing too much work for now on the front end of the application, since a lot might change.

You could try moving the logs into tmpfs
Is this the cause of some of the slowdown? I wasn't aware of any issues with the logs. Once I get everything stable I usually turn the logs to "none" in the core settings.

You could still give it a go. I think most of the slowdown is caused by the amount of devices, but worth a try since your logs are already minimal - no important info would be lost.

Are there any plans to make any kind of results/summary page based on subnets rather than having them all listed at once?

I thought briefly about it in the past - maybe. You could also think about running multiple NAX instances and then syncing the data into one. You could then do most administrative tasks in the agent nodes. Then, if using graphana, have a graphana dashboard for all the data synced from the main hub node. See:

• https://github.com/jokob-sk/NetAlertX/tree/main/front/plugins/sync/
• https://jokob-sk.github.io/NetAlertX/API_METRICS/

I'm unsure if that is what "Network" does, as it never loads for me. I assumed because it is trying to render thousands of devices.

It should be a visual representation of your network, but can;t imagine this being useful with that many devices. Previously I tested the app with 3000 devices and it was manageable - but even then I didn't test this section: https://jokob-sk.github.io/NetAlertX/NETWORK_TREE/?h=netwo

[adamoutler]

I asked Gemini about a hunch. I figure this might be worth dropping here.

I'm not sure if supporting other databases might help or not. But Gemini seems to think so.

[adamoutler]

There's no getting around the amount of time it takes to query 4000 devices, but we can work on the in-between and potentially parallelization.

[jokob-sk]

Would definitely help, just blocked by a bunch of other things. Would need to refactor lots of things.

#1141

[jokob-sk]

Long term goal -> closing for now