Server CPU Usage

[ASiC666 6]

Hi,

Can I suggest a new metric on the servers list?
Would it be possible to add a server CPU usage % please?

I might be wrong but I believe in some cases that although there seems to be bandwitdh available, the CPU is hammered (probably IRQs) as the actual tunnel speed is very low.
For example I am on a server with ~250mb to spare (according to the servers page) and I can barely do ~2.5MB/sec on either well seeded torrents or usenet.

Thanks

 

[Tech Jedi Alex 1570]

Before suggesting such a change for everyone, did you try changing up connection settings on your end so that you may know for sure it's the server and not you? Servers, ports, protocols? MTU? sndbuf/rcvbuf? uTP and limits in torrent client? Stuff like that.

[ASiC666 6]

On 8/30/2026 at 10:02 AM, Tech Jedi Alex said:

Before suggesting such a change for everyone, did you try changing up connection settings on your end so that you may know for sure it's the server and not you? Servers, ports, protocols? MTU? sndbuf/rcvbuf? uTP and limits in torrent client? Stuff like that.

I have not suggested a change; I've suggested a new feature.
Some more transparency to airvpn backend, if you like.

And yup, I have attempted everything in the shopping list. It is definitely not me...

You don't have to be defensive; I'm not accusing airvpn.
I expect from a service that it does not impose limitations per user, to be abused here and there.
And I will gladly accept the above fact to any forced limitations, censorship etc

My thought process for the suggestion was to have another metric in the mix for selecting a desired server
Thats it.
 

[Tech Jedi Alex 1570]

6 hours ago, ASiC666 said:

I have not suggested a change; I've suggested a new feature.

You've suggested a new feature which necessitates changing the existing configuration of a plethora of tools: rrdtool for the graphs, the status webpage (if it should be displayed in the overview), webpages for all servers, the AirVPN API. These changes must then be incorporated into Eddie, Eddie for Android and the AirVPN Suite; after all, what's a rrdtool graph a human uses to determine the "best" server if the tooling does not use the data it is based on? Because if it's missing, I'll bet you that tickets and forum topics will emerge asking to incorporate this into the clients' server choice and API output. And while we're there, why not also publish RAM usage, they will ask..
You are trying to delegate engineering a "solution" to a problem you are having to someone who needs to consider the implications of its implementation on the system as a whole.

Whether the CPU is the problem can easily be gauged by yourself by lowering the cipher to AES-128, which AirVPN supports, or using ChaCha20-Poly1305 altogether which is said to be easier on the resources with similar strength to AES. Did you try that? Because you write:

6 hours ago, ASiC666 said:

And yup, I have attempted everything in the shopping list

and I'm not sure you even thought of changing the cipher. Hard to tell what this shopping list contains; what I am seeing is a feature request stemming from a problem X you seemingly can't solve, but think Y can solve the problem, so you suggest Y but Y needs work from someone else, massive tests, rollouts, bug fixing,..
 

6 hours ago, ASiC666 said:

It is definitely not me...

It is most definitely only you. There are people transferring with 20+ MB/s right now, some are even beyond a gigabit (>110 MB/s, which is kinda awesome. And makes me slightly jealous somehow).
 

6 hours ago, ASiC666 said:

I expect from a service that it does not impose limitations per user, to be abused here and there.

Valid expectation; that's why you're with AirVPN, right? Because there are no such limitations, obviously. See the top users above.
 

6 hours ago, ASiC666 said:

My thought process for the suggestion was to have another metric in the mix for selecting a desired server

I'm defensive because of that thought process: Why a feature request? Why not a help request? If you already opened one without a resolution for you, I apologize, please link to the thread. But if you don't give enough info, people can and will misunderstand..

[Staff 10636]

1 hour ago, Tech Jedi Alex said:

And makes me slightly jealous somehow).

Hi,

not to make you even more jealous but look at what we caught on Aug 26: 😋

[ASiC666 6]

2 hours ago, Tech Jedi Alex said:

You've suggested a new feature which necessitates changing the existing configuration of a plethora of tools: rrdtool for the graphs, the status webpage (if it should be displayed in the overview), webpages for all servers, the AirVPN API. These changes must then be incorporated into Eddie, Eddie for Android and the AirVPN Suite; after all, what's a rrdtool graph a human uses to determine the "best" server if the tooling does not use the data it is based on? Because if it's missing, I'll bet you that tickets and forum topics will emerge asking to incorporate this into the clients' server choice and API output. And while we're there, why not also publish RAM usage, they will ask..
You are trying to delegate engineering a "solution" to a problem you are having to someone who needs to consider the implications of its implementation on the system as a whole.

Whether the CPU is the problem can easily be gauged by yourself by lowering the cipher to AES-128, which AirVPN supports, or using ChaCha20-Poly1305 altogether which is said to be easier on the resources with similar strength to AES. Did you try that? Because you write: and I'm not sure you even thought of changing the cipher. Hard to tell what this shopping list contains; what I am seeing is a feature request stemming from a problem X you seemingly can't solve, but think Y can solve the problem, so you suggest Y but Y needs work from someone else, massive tests, rollouts, bug fixing,..
 
It is most definitely only you. There are people transferring with 20+ MB/s right now, some are even beyond a gigabit (>110 MB/s, which is kinda awesome. And makes me slightly jealous somehow).
 
Valid expectation; that's why you're with AirVPN, right? Because there are no such limitations, obviously. See the top users above.
 
I'm defensive because of that thought process: Why a feature request? Why not a help request? If you already opened one without a resolution for you, I apologize, please link to the thread. But if you don't give enough info, people can and will misunderstand..

Yes, I am aware of what is involved if a suggestion, feature or a change gets approved and passes to R&D stage, then to system testing, then to UAT, release cycles etc etc.
But before all these, it is just a suggestion. A conversation piece, if you like.

Apologies for the "shopping list" bit; I was referring to the "Servers, ports, protocols? MTU? sndbuf/rcvbuf? uTP and limits in torrent client? Stuff like that." in your initial reply, regrettably without finesse.

So all though you seem sure that I haven't done any testing with different configurations, I can gurantee you that I have. 
I do not expect you to take my word on this. 
However, I was expecting a discussion with suggestiong thrown in, fresh ideas etc.

What we've got ourselves into is an aggressive exchange; not very constructive

On the technical aspect now, my goto cipher is ChaCha20-Poly1305 over WireGuard. 
I use gluetun in a couple docker containers and also two client connections from my UDM Pro, one routing selected devices and the second to specific internet endpoints using policy-based routes.
The MTU is set by default to 1380 and lowering when testing.

I attempted to throw the idea of server metrics because the issues I am having seem to be server specific:
The speed is not ideal on server X but it is better on server Y in the same region and similar bandwitdh usage.

From reading the forums, I can see a number of people are facing similar issues

The top user speed tables are impressive but IMHO they lack the context of how the users are achieving these speeds (what cipher/service/number of concurrent connections/destination etc). Obviously I am not expecting any table to provide such info!

I have opened a couple of tickets in the past mostly around time critcal applications such as voip and iptv. I believe under a different username, it has been some time.
In the end of the day, I am not too fussed about these issues as I mostly use airvpn for file transfers in which temperamental speeds are not the end of the world!

Anyway, my cat has started destoying the sofa as I'm late for her dinner! 
Cheers 


 

[Staff 10636]

On 8/28/2026 at 12:30 PM, ASiC666 said:

Would it be possible to add a server CPU usage % please?

Hello!

It is possible, of course, but could it be misleading? As you know, a server’s overall load doesn’t tell you much on its own. We’re proud to offer a level of transparency that is unmatched in the industry. In fact, we can’t recall seeing another public server monitoring page in the past 15 years or so that shows anything remotely comparable to ours. Thus we will consider your suggestion seriously, but it must be pondered carefully.

Kind regards
 

[ASiC666 6]

1 minute ago, Staff said:

Hello!

It is possible, of course, but could it be misleading? As you know, a generic load on a server does not tell much by itself. We're proud to offer more transparency than any other service: at the moment we can't recall any "server monitor" vaguely comparable to ours in the last 15 years or so. Thus we will consider your suggestion seriously. It must be pondered carefully.

Kind regards
 

Hiya,

Yeah, you are spot on about being a misleading metric, especially with the missing context on how and where these CPU cycles are consumed.

There is no question about airvpn's transparency. You have every right to be proud 

[oassQ9w4cbl4AySZhhth%p36x 9]

what i would prefer to see instead (and you already do 90% of this anyway so it wouldn't be a huge addition) is some kind of relative heuristics on the 'Latency from locations, IPv4' selection

im assuming constant pings are being run on all those locations anyway to get those values and if you log and display the current, 7d and 30d averages - then its easy to see at a glance 'oh shit this servers ping is up massively, it must be getting taxed or routing is dodgy'

thats the kind of an 'at a glance' info i would like to see, not cpu - it could be a cpu problem but ultimately none of us give a shit about the cpu but about how the network is functioning - so greater network related observability would be really helpful as an identification metric of 'look into this' 

[Tech Jedi Alex 1570]

2 minutes ago, oassQ9w4cbl4AySZhhth%p36x said:

then its easy to see at a glance 'oh shit this servers ping is up massively, it must be getting taxed or routing is dodgy'

That is what the packet loss metric is for. It is the absolute indicator on whether the server line is troubled and so definitely causes problems for everyone. With that said, it's only shown if it's above 20%, as a server problem in the status log, but 10% packet loss can equally be destructive to some experiences, and 10% is not shown. In the API, for <20% the health key will still value ok, which is actually not quite ok. I mean, if you stream something with 10% packet loss via VPN, you can be sure that DASH streams at least will be inconsistent and wildly change quality.
My vote is for that metric to be shown, also in the API – if I had a vote.
 

On 8/31/2026 at 8:34 PM, ASiC666 said:

However, I was expecting a discussion with suggestiong thrown in, fresh ideas etc.

What we've got ourselves into is an aggressive exchange; not very constructive

Apologies. The majority of such feature requests are made by people who gave up troubleshooting somewhere, not exhausting all options. Most of all, the option of asking for help. Once their own (limited) repertoire of ideas runs dry, it must be the server, they conclude – while others still manage to max out their line on that same server. Must be the server, it's definitely not me, I tried everything under the sun. Probably only under their sun.
And you can take my word for it as someone who roamed these forums for, I believe it's 13 years, now. It's a repeating leitmotif. Might've worded it less confrontationally, I admit. 🙏