Servers power up shown in the web monitor

[bnrrteterstnjrsj45 0]

Hello, as I understand in Eddie client at servers load stat we seeing only downstream? Can this consider upstream too? Or, maybe, both separated.

Thanks.

[Staff 10114]

Hello!

The total availability of bandwidth is "full duplex" (so a 1 Gbit/s server can provide up to 2 Gbit/s) but the "used" bandwidth is the sum of up+down, combined. You can safely assume, due to how a VPN works, a near perfect symmetry between upload and download, as each incoming (from the Internet) packet must be sent out to the client, and vice-versa.

Kind regards
 

[SurprisedItWorks 52]

So on your servers page, a report of 300 / 1000 actually means 150ish up, 150ish down, out of a theoretical maximum of 1000up + 1000 down?  If so, that "/" character is ambiguous, being linguistic rather than mathematical. Better to have both numbers in the same units, either unidirectional or total. 

[Staff 10114]

@SurprisedItWorks

Hello!

Yes, it's inconsistent at the moment.

Kind regards
 

[Staff 10114]

  On 6/27/2023 at 5:57 PM, Staff said:

@SurprisedItWorks

Hello!

Yes, it's inconsistent at the moment.

Kind regards
 

@SurprisedItWorks

Hello!

Resolved, please see also https://airvpn.org/forums/topic/56492-congested-canadian-servers/?do=findComment&comment=224442

Kind regards
 

[arteryshelby 25]

Wouldnt it be better to stick to fullduplex throuput? One Server with 1G can push 1G to a client, not 2G, i mean.. yes.. if you account it as halfduplex, but thats a really odd messurement to use IMHO.


 

[Staff 10114]

  On 7/15/2023 at 2:09 PM, arteryshelby said:

Wouldnt it be better to stick to fullduplex throuput? One Server with 1G can push 1G to a client, not 2G, i mean.. yes.. if you account it as halfduplex, but thats a really odd messurement to use IMHO.

Hello!

No, it wouldn't be better. Only now the measurement is fully consistent. This measurement system shows you things as they are. The old system added up+down bandwidth of clients, but cut in half the server availability, therefore a server with 1000 Mbit/s in use would look at capacity while in reality it can still give another 1000 Mbit/s - you should see by yourself that this system can't hold.

If we kept the old system we would keep receiving (and rightly so) a lot of complaints for servers which seem at capacity or overcapacity while they are not. If we cut the used bandwidth in half, we would rely exclusively on the VPN symmetry, which is not truthful when a server must face floods or non-VPN traffic in general.

This is again a community driven request and in this case we share the objections against the old inconsistent system, which made sense when CPU and other bottlenecks capped the server, or when we had 500 Mbit/s full duplex lines.

Kind regards
 

[arteryshelby 25]

Thanks that you take care of the feedback.

Why not just display fullduplex as maximum (e.g. 1G/10G) on Serverstats but account for just halfduplex the traffic (e.g. only counting inbound or outbound, should not really matter)? This would still make it consistent while maintining a more real-world like view.

 

[Staff 10114]

  On 7/15/2023 at 6:21 PM, arteryshelby said:

Why not just display fullduplex as maximum (e.g. 1G/10G) on Serverstats but account for just halfduplex the traffic (e.g. only counting inbound or outbound, should not really matter)? This would still make it consistent while maintining a more real-world like view.
 

Hello!

Sorry, we can't understand you, we just explained why that would not be truthful.

Kind regards
 

[arteryshelby 25]

We are bassicly talking about customers without a little network knowledge right? because if you have knowledge, you know how airvpn accounts its traffic so the problem you describe / the reason the change was made, is not there in the first place.

If i know nothing about networking and i see a vpn which says "2000 mbit" i would assume that i can ues this 2000 mbit. But this is not the case.

for vpn/forwarding based system the asymmetry due to DoS(/ Downloading stuff on server / whatever) is such an edgecase that its still much closer to realworld view to just cut the used bandwith in half then showing halfduplex as maximum.
If you want to account for download / dos, it would also be possible to use inbound usage as server usage? Outbound should never really be greater then inbound in this usecase? Or just take whatever direction has more load.

[Staff 10114]

  On 7/15/2023 at 7:12 PM, arteryshelby said:

If i know nothing about networking and i see a vpn which says "2000 mbit" i would assume that i can ues this 2000 mbit. But this is not the case.

Hello!

If you know nothing about networking you would probably ask for information or get informed and then you discover that this is exactly the case.

Kind regards
 

[Staff 10114]

Hello!

We have merged two threads into this one so you can continue here the discussion and leave the "News" forum more readable.

Kind regards
 

[arteryshelby 25]

  On 7/15/2023 at 7:17 PM, Staff said:

Hello!

If you know nothing about networking you would probably ask for information or get informed and then you discover that this is exactly the case.

Kind regards
 

I dont get it.. People ARE asking for information and you state multiple times how the accounting works. If you want to honor your own statement - keep it that way?

Displaying halfduplex as maximum while ALL OTHER vpn companies display fullduplex is just wrong. I get that you also account traffic twice so the load is correct, but the maximum is wrong.. As a customer i expect 2 gbit download speed from 2 gbit maximum server. Not 2 gbit halfduplex.

I provided various solutions for mentioned problems that would not require the workarround with displaying halfduplex capacity as maximum.

PS: i know that im hard on you. I do this because you are (mostly) open for my harsh feedback to provide the best possible service.

[Staff 10114]

  On 7/15/2023 at 7:25 PM, arteryshelby said:

Displaying halfduplex as maximum while ALL OTHER vpn companies display fullduplex is just wrong.
 

No, it's just right, because it shows the maximum capacity of a server line. Can we see the real time servers monitor of other companies showing what you claim?
 

  Quote

I get that you also account traffic twice so the load is correct, but the maximum is wrong..

No, we don't account traffic twice, otherwise the load would not be correct. We show the exact throughput (sampled every other minute) on the server. For the top throughput of clients please consult the proper table.
 

  Quote

As a customer i expect 2 gbit download speed from 2 gbit maximum server.

The real time servers monitor shows the server status, not your expectations.
 

  Quote

I provided various solutions for mentioned problems that would not require the workarround with displaying halfduplex capacity as maximum.

Your solutions would display false data, while the real time servers monitor displays true data within the boundary of the sampling frequency.

Kind regards
 

[arteryshelby 25]

  On 7/15/2023 at 7:34 PM, Staff said:

No, it's just right, because it shows the maximum capacity of a server line. Can we see the real time servers monitor of other companies showing what you claim?
 
 

I guess we have a different view on things in this case. Azirevpn does it correcty, IVPN has no used bandwith in mbits but shows load in % and uses fullduplex as maximum (nvm - no info) . Mullvad shows fullduplex maximum.. the list goes on.. Name on provider that uses your solution
 

No, we don't account traffic twice, otherwise the load would not be correct. 

The phrasing was wrong, but i we ment the same.
 

  Quote

The real time servers monitor shows the server status, not your expectations.

I still argue that displaying halfduplex as maximum is not technical correct either.
 

  Quote

Your solutions would display false data

Well it depends on what of the two solutions we are talking about:
Adding download + upload and cutting this in half would technically not be correct.. (even tho accurate enough IMHO)
but if you take the greater value (inboud or outbound) and use this as system usage, it would not be possible to use the difference of available bandwith between my accounting any your accounting (adding download + upload seperatly) anway because the line would be congested in either the one directions before the addiotional difference in inbound/outbound can be used. So.. does this really matter? I would say this technically can be seen as the system load.

Take this example:
A server has 200 mbit outbound and 800 mbit inbound usage
with your accounting the usage would be 200 + 800 = 1000 from 2000 (halfduplex)
with my account the usage would be 800 (greater value) from 1000 (fullduplex)

even tho we see a difference in server usage - 800 and 1000 (which you might see as "wrong data") the availabile bandwith for customers would still only be 200 mbit (fullduplex) or 400 mbit (halfduplex), but with your accounting it would look like 1000 mbit (which in halfduplex terms is still available.. but not for the usecase as vpn server). I would say my accounting is more relevant from a customers perspective.

[Staff 10114]

  On 7/15/2023 at 7:47 PM, arteryshelby said:

I guess we have a different view on things in this case. Azirevpn does it correcty, IVPN has no used bandwith in mbits but shows load in % and uses fullduplex as maximum. Mullvad shows fullduplex maximum.. the list goes on..

Hello!

AzireVPN uses our new method, we see now from their stats. They show bandwidth in and bandwidth out of each server. They don't have graphs, stats by time periods and a ton of other features we offer, but they still show up + down ("in" + "out" in their monitor).

Note how it's not symmetrical, showing once again that your solution was indeed inapplicable. For example in this moment they declare that a server in France uses 1200 Mbit/s "in" and 109 Mbit/s "out". They also write: "Each server is connected with two 1 Gbit/s links towards the switch."

In iVPN,, oh well... that would be a server monitor in your opinion?!  It's a list of servers with a percentage, no stats, no graphs, no history, no nothing...
 

  Quote

Name on provider that uses your solution

We can't see any provider offering a server monitor like ours. To the best of our knowledge this is a very exclusive feature of AirVPN. So we can't name any simply because nobody has it. But Azire (you mentioned it) adopts the same solution by publishing "in" and "out" flow.
 

  Quote

Adding download + upload and cutting this in half would technically not be correct.. (even tho accurate enough IMHO)

As shown by your own Azire example, it wouldn't be accurate.
 

  Quote

but if you take the greater value (inboud or outbound) and use this as system usage, it would not be possible to use the difference of available bandwith between my accounting any your accounting (adding download + upload seperatly) anway because the line would be congested in either the one directions before the addiotional difference in inbound/outbound can be used. So.. does this really matter?

That's of course false. You may have asymmetries which make this quoted statement false. Even a scriptkiddie modest flood counts significantly, in this case, either inside the VPN, or incoming from outside the VPN itself.

Publishing raw data as they are is the most reliable way, it's the way recommended by an important amount of community members, and it does not require data manipulation. By clicking the name of the server in our real time server monitor you can still see the distinction of bandwidth "up" and "down" of course, that feature, together with the graphs, remain.

Kind regards
 

[arteryshelby 25]

  On 7/15/2023 at 8:27 PM, Staff said:

Hello!

AzireVPN uses our new method, we see now from their stats. They show bandwidth in and bandwidth out of each server. They don't have graphs, stats by time periods and a ton of other features we offer, but they still show up + down ("in" + "out" in their monitor).

Note how it's not symmetrical, showing once again that your solution was indeed inapplicable. For example in this moment they declare that a server in France uses 1200 Mbit/s "in" and 109 Mbit/s "out". In iVPN,, oh well... that would be a server monitor in your opinion?!  It's a list of servers with a percentage, no stats, no graphs, no history, no nothing...
 
 We can't see any provider offering a server monitor like ours. To the best of our knowledge this is a very exclusive feature of AirVPN. So we can't name any simply because nobody has it. But Azire (you mentioned it) adopts the same solution by publishing "in" and "out" flow.
 

I indeed had a few things wrong in my mind here, silly brian. There is nothing wrong with display in and out seperatly. In fact i higly apprechate that. But adding theese two up and using halfduplex as maximum does not represent accurate available bandwith / serverload for customers but total avabile halfduplex bandwith which customers do not care about (see my example reagarding asymmetrical load case).

 

  Quote

That's of course false. You may have asymmetries which make this quoted statement false. Even a scriptkiddie modest flood counts significantly, in this case, either inside the VPN, or incoming from outside the VPN itself.

You understood wrong what i meant. I edited my reply and added an example. Perhaps things are clearer now.

[Staff 10114]

  On 7/15/2023 at 8:32 PM, arteryshelby said:

I indeed had a few things wrong in my mind here, silly brian. There is nothing wrong with display in and out seperatly. In fact i higly apprechate that. But adding theese two up and using halfduplex as maximum does not represent accurate available bandwith for customers but total avabile halfduplex bandwith which customers do not care about (see my example reagarding asymmetrical load case).

Hello!

OK, and please let the community help us understand what customers care about, and let us decide accordingly, especially when hundreds of them asked for this modification.

Kind regards
 

[arteryshelby 25]

well.. most see the inconsitence between load and total available bandwith, THIS was indeed a problem which you tried to fix. Just because i say your fix is wrong (from my perspective) doesnt mean the community is wrong that a fix is needed in gernal  and doesnt indicate that hundreds are also against my view

[revsplus 8]

  On 7/15/2023 at 7:12 PM, arteryshelby said:

We are bassicly talking about customers without a little network knowledge right? because if you have knowledge, you know how airvpn accounts its traffic so the problem you describe / the reason the change was made, is not there in the first place.

If i know nothing about networking and i see a vpn which says "2000 mbit" i would assume that i can ues this 2000 mbit. But this is not the case.

for vpn/forwarding based system the asymmetry due to DoS(/ Downloading stuff on server / whatever) is such an edgecase that its still much closer to realworld view to just cut the used bandwith in half then showing halfduplex as maximum.
If you want to account for download / dos, it would also be possible to use inbound usage as server usage? Outbound should never really be greater then inbound in this usecase? Or just take whatever direction has more load.

Hi,
I see some confusion here that probably leads to dialogue problems. Please note that a 2 Gbit/s half duplex "line" has nothing to do with a 1 Gbit/s full duplex one. In most real life use cases the latter is preferred and is superior for obvious reasons. To understand the difference just read one of the many technically excellent articles you can find on the web.

I don't see why you discern between Azire and AirVPN in this case. Azire declares a maximum of 2 Gbit/s per server with "two 1 Gbit/s links", while AirVPN declares the same with 10 Gbit/s or 1 Gbit/s "full duplex" lines (*). In the first case, Azire hints to a dual simplex communication, which colloquially is anyway named "full duplex", while AirVPN says "full duplex", hinting to a real, not emulated full duplex. Both definitions are technically acceptable, if not rigorous, and are equivalently effective for a collision free environment, which is what we absolutely need when connecting to a VPN.

The previous error of AirVPN was claiming a total of 1 Gbit/s bandwidth even on servers with 1 or 10 Gbit/s full duplex comms. As far as I understand this was a "historical" error born when the hardware bottlenecks or the links could not allow the server to beat 1 Gbit/s anyway, so it went unnoticed even after the migration to 1 and 10 Gbit/s full duplex comms. Now keeping this error would be masochistic, I guess, besides being a technical error, because that limit is no more for various improvements and for WireGuard.

(*) "Staff" has mentioned several times that Dallas and Alblasserdam servers are connected with 10 Gbit/s full duplex lines, with 10 or 11 servers per line and each server sharing the total bandwidth.

last minute add: on a less serious note, comparing the AirVPN server monitor with iVPN or Azire ones is offensive for AirVPN I guess, so you might have offended someone 😂

[fsy 37]

@Staff
@arteryshelby
@revsplus

Mullvad does the same. It publishes 20 Gbit/s for 10 Gbit/s full duplex l., 2 Gbit/s for 1 Gbit/s full duplex and 1 Gbit/s for 500 Mbit/s l. It's normal and appropriate to me. AirVPN finally patches the "inconsistency" affecting the monitor ever since... always.

Quick OT: Mullvad shrunk recently the amount of servers. From 769 servers a few months ago, to 666 (Holy Moly, the number of the Beast!!!). Just seen today. Wondering whether it comes from the deletion of port forwarding...

Disclaimer: I don't want to compare Mullvad embryonic and undeveloped monitor with AirVPN interstellar monitor, me don't want to offend nobody. 😁

[arteryshelby 25]

  On 7/16/2023 at 8:41 AM, revsplus said:

Hi,
I see some confusion here that probably leads to dialogue problems. Please note that a 2 Gbit/s half duplex "line" has nothing to do with a 1 Gbit/s full duplex one. In most real life use cases the latter is preferred and is superior for obvious reasons. To understand the difference just read one of the many technically excellent articles you can find on the web.

I don't see why you discern between Azire and AirVPN in this case. Azire declares a maximum of 2 Gbit/s per server with "two 1 Gbit/s links", while AirVPN declares the same with 10 Gbit/s or 1 Gbit/s "full duplex" lines
 

The difference is that Azire acutal has 2x1Gbit fullduplex lines, e.g. 4Gbit in Airvpn terms. Two phyiscal 1G fullduplex lines.. this is my hole point why im arguing.. Also Air doesnt display 10 Gbit fullduplex anymore with 20 gbit maximum serverload for 10 gbit systems.. this is halfduplex capcity.

[arteryshelby 25]

  On 7/16/2023 at 9:53 AM, fsy said:

@Staff
@arteryshelby
@revsplus

Mullvad does the same. It publishes 20 Gbit/s for 10 Gbit/s full duplex l., 2 Gbit/s for 1 Gbit/s full duplex and 1 Gbit/s for 500 Mbit/s l. It's normal and appropriate to me. AirVPN finally patches the "inconsistency" affecting the monitor ever since... always.
 

No they dont

I can open up a ticket with mullvad if you want offical reply. But looking at the server list there a quite a few 1 and 10gbit servers, that wouldnt make sense otherwise.

I understand that Air tries to patch the inconsistency, but the solution is .. Stick to fullduplex pls.

[revsplus 8]

  On 7/16/2023 at 11:26 AM, arteryshelby said:

The difference is that Azire acutal has 2x1Gbit fullduplex lines, e.g. 4Gbit in Airvpn terms. Two phyiscal 1G fullduplex lines.. this is my hole point why im arguing.. Also Air doesnt display 10 Gbit fullduplex anymore with 20 gbit maximum serverload for 10 gbit systems.. this is halfduplex capcity.

Hi,
I wouldn't say so. Azire writes very clearly "The switch is connected to one or two 10 Gbit/s uplinks which can be upgraded to 40 Gbit/s QSFP uplinks when needed. Each server is connected with two 1 Gbit/s links towards the switch."
To use a similar terminology in AirVPN datacenter for which this info has been made public, for example Alblasserdam, the servers are connected with 4 (four) or 2 (two) 1 Gbit/s links, or 2 (two) 10 Gbit/s links to 40 Gbit/s switches/uplinks and (two? I don't remember) high speed routers towards AMS-IX. You overlap or confuse different things and you stubbornly remark half duplex once more... I'm afraid I can't help anymore here, have a nice day.

@fsy
 

  Quote

Quick OT: Mullvad shrunk recently the amount of servers. From 769 servers a few months ago, to 666 (Holy Moly, the number of the Beast!!!). Just seen today. Wondering whether it comes from the deletion of port forwarding...

Really?! Who knows. I am one of those who left Mullvad when they decided to drop port forwarding (I have been refunded) - anyway I was always an old customer of AirVPN - and by hearsay thousands have migrated from Mullvad. However it could be a combination of upgrades and more competition (Google VPN? though I'm not sure they are serious competitors for Mullvad or AirVPN target).

The recent liquidity problems of other VPNs you wrote about in a message some time ago, causing sales and bankrupts, on top of the remarkable, supposed migration (if true) make your question legitimate: maybe the sector of independent VPNs started to suffer because of big actors. After all 103 servers less in a short time are not a negligible amount. We'll see, at the moment I am more than happy again with AirVPN only (if this quality stands I don't need any other VPN atm). Have a nice day!

 

[arteryshelby 25]

  On 7/16/2023 at 2:09 PM, revsplus said:

Hi,
I wouldn't say so. Azire writes very clearly "The switch is connected to one or two 10 Gbit/s uplinks which can be upgraded to 40 Gbit/s QSFP uplinks when needed. Each server is connected with two 1 Gbit/s links towards the switch."
To use a similar terminology in AirVPN datacenter for which this info has been made public, for example Alblasserdam, the servers are connected with 4 (four) or 2 (two) 1 Gbit/s links, or 2 (two) 10 Gbit/s links to 40 Gbit/s switches/uplinks and (two? I don't remember) high speed routers towards AMS-IX. You overlap or confuse different things and you stubbornly remark half duplex once more... I'm afraid I can't help anymore here, have a nice day.

Lets keep uplinks out of the picture. They for sure are shared like in most envoriments but aslong as the overcommit isnt to bad that shouldnt be problem.

Im not overlapping anything here. Air is adding up+download as total server capacity while Azire displays acually physical links like its standard in every datacenter enviroment.
If i order a 20gbit port from a datacenter, its _always_ fullduplex, e.g. 2x10gb (or in Air terms.. 40 Gbit 😂).. not 1x10Gbit which might also add up to 20 gbit inbound/outbound combined LOL.

Look at my example above and tell me that you think Airs way of calculating isnt total bonkers:

  Quote

Take this example:
A server has 200 mbit outbound and 800 mbit inbound usage
with your accounting the usage would be 200 + 800 = 1000 from 2000 (halfduplex)
with my account the usage would be 800 (greater value) from 1000 (fullduplex)

Air would calculate the system as 50% serverload with 1000 mbit availible to clients, while in reality only 200 mbit could be used for forwading since the interface would otherwise be completly congested in inbound direction.

With my calculation you see actuall available forwarding capacity to clients.
While Air shows total available accross inbound/outbound bandwith combined which might only be available in either of the one interface directions and is completly useless for clients as a usecase for forwading traffic (e.g. using as vpn exit).