About system requirements, what is considered to be "moderate traffic"?

The countly docs suggest the following:

  • At least 2 GB of RAM for testing purposes
  • 4 GB of RAM for moderate traffic
  • 16-64 GB RAM or more for high traffic servers

My question is, what is considered "moderate traffic"?
A ball park estimate in numbers would make more sense to me.

You are 100% right @FelDev - I will modify that part so it reads better.

We are also working on a page where you can easily guess what kind of a server you would need, depending on the traffic.

If you could let me know your very rough traffic, I would be happy to provide the server sizing :slight_smile:

1 Like

Awesome! Thanks.

Here are screenshots of the traffic our Countly server received in the past 24 hours. (In 2 different posts because I'm a new user...)

For people who can't read images:

Max Avg Last
Private Outbound 21.68 b/s 542.24 mb/s 0.00 b/s
Private Inbound 818.84 b/s 207.31 b/s 414.21 b/s
Public Outbound 7.56 Mb/s 3.05 Mb/s 3.24 Mb/s
Public Inbound 26.56 Mb/s 11.30 Mb/s 9.64 Mb/s
In Out Combined
Total Traffic 164.35 GB 44.08 GB 208.43 GB


For people who can't read images:

Max Avg Last
Private Outbound 249.03 b/s 4.05 mb/s 3.05 b/s
Private Inbound 76.22 b/s 512.23 b/s 69.08 b/s
Public Outbound 2.27 Mb/s 1.03 Mb/s 993.74 Mb/s
Public Inbound 17.05 Mb/s 3.92 Mb/s 3.34 Mb/s
In Out Combined
Total Traffic 42.28 GB 11.11 GB 53.40 GB

Unfortunately I am not sure about in and out traffic. Do you mind whether you can go to dashboard and under datapoints, what does it say for this month and last month please? If you think it is private information, feel free to send it along to gc@count.ly please.

Of course!

Our actual numbers will be bigger than this because our Countly server has been down for about 8 days in September and about 10 in October as we're still figuring out how to use the tool, but here are our numbers anyways:



As for our current system resources, we first used Countly in production on a standard Linode with 2GB of RAM, eventually it completely crashed.

We upgraded the linode to 4GB of RAM and 2 CPU, eventually that crashed too.

So, we upgraded the linode to 8GB of RAM and 4 CPU. It did not crash, but we would still get alerts from Linode about CPU usage and inbound traffic rates.

That's when I searched the docs and saw that

Running your Countly instance on a dedicated or barebone server greatly increases network throughput and CPU performance.

So, now we run this:

  • 8GB of RAM
  • 4 CPU
  • 160GB of storage
  • On a dedicated CPU

No crashes to report yet, but I got mail from Linode informing me that the CPU was being used heavily and inbound traffic rates were high ("averaging 14.53 Mb/s for the last 2 hours").

Being used to PAAS like Heroku or even easier hosting with services like Netlify, I feel like I'm just taking shots in the dark with this server management thing...

End of the novel :sweat_smile:

Checking your datapoints, it is 50M datapoints for September itself, plus it has been down for some days, meaning you are looking at 75M datapoints monthly. To be on the safe side we can consider that 100M datapoints / mo (which is over-moderate-traffic). Our basic assumption is (roughly) 1 CPU and 2GB or RAM per 5M datapoints on shared instances, hence I suggest to go with 16 CPU and 32GB of RAM at a minimum. Since you are on a dedicated CPU, you might be good but I still find it very risky to run 100M DP on 4CPUs. :slight_smile:

:heart_eyes: Numbers :heart_eyes:


Well thanks, this is very helpful.

Cheers and have fun with Countly :star_struck: !

1 Like

For posterity...

The server setup that ended up working pretty well for us is the following:
32 GBs of RAM
8 CPUs
640GB of storage (non-SSD)
On a non-dedicated server.

It seems like 2GB per CPU was not enough.
I think it's because our DB is hosted on the same server (standalone). (Countly docs on the subject)

Then it started working really well once I added an index on one often queried collection.

Hope this helps someone sometime!

Great! Many thanks letting us know. Datapoints thing has a lot of different setups, or patterns let me say, based on whether your users are evenly distributed or not definitely. Your comments have given us just another food for thought.


1 Like