More Canadian Servers

[p9974839 2]

Been noticing that many of your Canadian servers are under increased load capacity (see snapshot). The 1 Gig servers are always listed at running at critical load capacity.
That being said, Will Air VPN be adding more Canadian servers in the near future?
Thanks for reading.

[Flx 76]

A move to 5 or 10 liners will be the smarter. It's 2023 and not even one 5 liner in US or Canada.
Only way this will happen is with Mr. @PJ approvel and/or someone from Air HQ.
Sadly...these complaints will continue...
Edit: @Clodo Does Eddie app have "multi-session" yet?

[OpenSourcerer 1359]

More than half of these servers have <50% load, so the country load is just under 50%. I do not understand why the attention is driven to the two reds in a sea of green and this is seen as a critical overload of the country.
Reconnect. Eddie and the DNS load balancer will select a server that's not red, and enjoy. Geez, guys.
 

On 2/5/2023 at 9:17 PM, Flx said:

Edit: @Clodo Does Eddie app have "multi-session" yet?

No one asked for it, so why do you expect that?

[Flx 76]

On 2/6/2023 at 3:39 PM, OpenSourcerer said:

No one asked for it, so why do you expect that?

Expected an answer from Mr. @Clodo...but this will have to "qualify" for now.

[knighthawk 19]

I have to agree with the OP in recent years CA servers are definitely more utilized. But I would say to the OP that it's not exactly an overall lack of servers or far higher user counts that drives some of those 100% instances, it's that air lets any user try to monopolize most of a box minus some reservations, and more people seem to be doing that and for longer periods.  In distant past when peoples connections were slower (50,100..even 200mbs) it was less of a thing, now I see stuff like below often with the advent of those blessed with gig connections at home or elsewhere.


Pic is from ~9pm today.. and I could tell by 6pm just surfing someone was hogging, its midnight now and that person is still at it, which is keeping the % at 100.  Hopefully when cases like that are going on by one session for long periods of time that sessions traffic is dynamically de-prioritized some how when load >90%, that or it's def time for some 10g boxes - though you'll just have the same issue again soon enough on 10g boxes. BTW that someone wants to pull 700+ for hours (everything but the reservation amount of 4 x usercount I assume)... I take little issue with that (in and of itself) _if_ and when there is little other pressure, but I would rather lose some ability or traffic priority at the extreme high end for my own traffic when the server is at or near 100% load in favor of other users pulling far less traffic during last sample period (5min avg?). Maybe that already happens to some extent dynamically, idk? Would love to understand the actual system in play when it comes to that if any, as it is I kinda assume we each get our min 4, then somewhat equally are competing over the rest, but hopefully it's actually more complicated then that.

On the just reconnect... ain't always that simple, some boxes are flagged by external parties in certain ways that prevent their use for certain things or traffic, while others are not and you never know what you'll get, not to mention config changes on the user end depending on scenario that can come into play, making it more of a pita then 'just reconnect forehead', but in some cases, maybe even a majority sure. 

-my worth less 2cents
 

[Staff 9751]

Hello!

34000 Mbit/s (full duplex, i.e. 68000 Mbit/s total) available in Canada, in peak hours about 23000 Mbit/s used. Plenty of bandwidth still available. Considering Canada extension, and the fact that WireGuard diffusion will cause higher bandwidth demand (mobile devices and routers can get more), the asymmetry between Vancouver (12000 Mbit/s)  and Toronto (52000 Mbit/s) will be progressively balanced according to needs (probably you have seen that 2000 Mbit/s have been added today in Toronto).

Please consider that the available bandwidth is real, it's not like having a 10 Gbit/s line, sharing it with dozens of servers, and declaring each of those servers as "10 Gbit/s" servers, as some competitors do.

@knighthawk

The kernel of the servers balance the demand quite efficiently. For example, if a client is draining 1 Gbit/s (i.e. 2 Gbit/s on the server, 100%) while 20 other clients are idle, the first client may continue to do so. As soon as the idle clients start demanding bandwidth, the provided bandwidth will be re-distributed in a very short time, and the initial client will be no more able to use 2 Gbit/s.

On top of that, we also take care (again thanks to kernel features) to distribute the incoming connections to different OpenVPN instances, so that a round robin distribution will contribute to keep balance. It's necessary only with OpenVPN, as a single OpenVPN process runs always in a single thread, while with WireGuard this is not necessary (and not applied).

Kind regards
 

[knighthawk 19]

@Staff

Thanks for answering that question about the dynamic redistribution upon noticing demand from previous idle\near idle sessions, very helpful information and explains why even on supposed 100% used system (at least in case when it's mostly from 1 or 2 sessions) I rarely have issues. 👍

[Flx 76]

This is the first time I've seen @Staff step-in to resolve an issue this fast. Congratz! :) 

[p9974839 2]

And it's getting worse as I am starting to see a "sea" of red and yellow.

[Staff 9751]

@p9974839

Hello,

we also see a sea of green. We still have 11000 Mbit/s free (full duplex), or if you prefer 22000 Mbit/s free total, in Canada which are never used by anyone, so the infrastructure is greatly oversized. Yellow means that you still have at least 1000 Mbit/s free on the server. Look at the very screenshot you sent us: it shows that you have plenty of choices in Canada on poorly used servers: 15 servers offer more than 1 Gbit/s (more than 500 Mbit/s full duplex)!

Check our previous message as well.

Kind regards
 

[OpenSourcerer 1359]

Also, to add to that, in the last few days I was using Haedus in the US which was always at 60-80% usage when I connected, and I did not experience degraded performance, mainly streaming and file uploads to a cloud. In my humble opinion you can treat any yellows as greens and simply use them.

These figures are momentary snapshots of the line's status, anyway. The reality is not so easy to measure, the status changes from second to second. Even the reds are usually temporary excesses in usage and don't depict prolonged overuse, see the rrd graphs on the server detail pages. Heck, even low-usage (~20%) servers can suddenly get a red status if someone with a gigabit link starts downloading seriously.

These screenshots are, again in my humble opinion, absolutely no proof for anything. If you want to proof continuous overuse of server bandwidth, start with the rrdgraphs.