All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Deliberately seeking a VPS with a somewhat bad connection
Hello LET!
Looking for something a bit unusual, I'm trying to find a test VPS that will let me test against actually bad network conditions (currently working on distributed networking software that must be highly reliable).
Before someone jumps into the comments and says it: yes, I am aware that tc and netem exist and can be used locally and/or on an actually reliable VPS, but they can't simulate things you would find in an actually awful network connection, like strange middleware boxes, bad routing choices, and more unusual types of packet loss / reordering. I want to discover those errors while developing and handle them there rather than in the many prod environments this code might eventually run in.
So the worse its networking is, the better it is. Not so bad that it's unusable (I still have to SSH in of course), but something at least more failure-prone than the average VPS is good (I can always stack with other tools server-side to add synthetic error on top of the natural error if needed). I'm in eastern North America, so the more latency and routing steps from me (and the more networking hardware in between me and the VPS), the better. Exotic locations, especially in Asia and/or Africa, would be preferred.
I would be mildly upset at greater than $5/mo for the VPS and would not consider more than $15/mo. Should have at least 10GB of disk, 512MB of RAM, an x86-64 or ARM CPU (no 32-bit), and ideally a high bandwidth cap (although not unlimited). IPv4 is optional (but would be a plus), I can work with just IPv6 if needed. Any ideas?
Comments
i don't read all this but i know @oloke will figure out something whatever it's bad or good
I would check out 3hcloud philippines vps.
https://3hcloud.com/
@zGato can probably give you some good recommendations.
I have used a lot of LET hosts and I have never used a provider that has a bad network in such a way
Most bad networks falls into either:
I've heard iHostArt is pretty dodgy, you can see their status page: https://uptime.ihostart.com/
justhost.ru specifically in Novosibirsk location.
Can't find more unreliable connection in production.
Deluxhost maybe?
Also adding in 4vps Korea and AU locations, they have bad routing due to single homed Cogent.
Instead of hunting for a bad VPS, check out Toxiproxy. It lets you inject exactly the kind of chaotic network conditions you want (packet loss, reordering, latency spikes, throttling, and other weird behavior) in any vps.
MassiveGRID for the shittiest network I've seen in a while:
DeluxHost most likely as well:
iHostART for the best "one day it works, next it doesn't" with a bunch of weird filters that make some things barely work.
A few of justhost.ru's RU locs are also quite unreliable, but depends on the upstream because connection to some services like Cloudflare or Google is quite stable.
Maybe someone will transfer you a vps on one of the hostsailor nodes where they can't figure out why it consistently shits the bed around 6-8am EST.
It was working fine for a while, but now the censorship is so awful that I'm going to move it to Moscow.
Quite accurate summary of all the suggestions so far (thanks to everyone who suggested, by the way!). I was kind of looking for something that had a spotty connection or dropped packets, but it seems the best I'll be able to find is bad peering and maybe some really nasty middleware hot garbage that messes with the box's connection in strange, unknowable, non-deterministic ways. I suppose if a VPS's connection was as awful as I am looking for, the provider probably wouldn't be in business for much longer...
Might just settle for the thing that's furthest from me in terms of number of network hops (justhost.ru seems like a potential winner).
I am aware of Toxiproxy and am already using it, happy to hear others are aware of it! I'm still searching for a VPS because even though I can try to cook up all kinds of weird settings for it and other network chaos monkeys, I still want to verify the numbers I'm picking are worse than reality for my theoretical worst-case situation (someone using my service with the most inconvenient routing that goes through the most trash network operators over a high latency connection).
@TrikeLike Honesty, you're going to want to simulate this on a network. Every case of "bad connection" is going to be different, and testing on one particular bad connection will give you very little insight into how to get around it because you don't have access to all the variables. Instead, you should be simulating various different real-life conditions. There is plenty of relevant material in literature.
I concur with @zgato. @MassiveGRID is the crappiest - and I've seen a few networks ...
Added bonus: the servers are crappy as well.
Any attempt to SSH directly into my iHostART VPSes results in a RST being sent, but I never bothered to diagnose why. I wonder if there's some filter there.
I've had better service buying through kyun.sh than directly from iHostArt
hostsailor is going into 2 or 3 weeks of issues, might be a good bet.
ha ha ha ha hahha, that was a good one.
It has been months.
About 12-15 months ago I had this kind of symptoms with VPS hosted in SkyLink's DC in Eygelshoven, and it was so bad that I canceled all nodes with all providers hosted there. I have no idea if it's still the case now.
HostSailor will absolutely not disappoint! They have VPSes that by now appear to be specially designed for this exact purpose.
I have both a mini sailor from Hostsailor for $8.98 and an ipv6 Ryzen 1c 1gb for €7 from Deluxhost that I would gladly get rid of .....
@TrikeLike how about Starlink? Too stable? Starlink+WiFi?
If I were to do this, it would be IPv6-only.
How much bandwidth usage do you have in mind?
Hey, for that kind of specific bad network testing, you could look into Time4VPS, or check Hivelocity and RamNode for their locations further away from you.
Is RamNode subpar these days? They used to be great a decade ago, I bought my first VPS ever from them in 2014 and I remember it being rock solid.
@slowservers you are one cool hosting provider! Love to see OpenBSD in the wild, might nab one of your servers for another project of mine at some point.
For bandwidth usage it doesn't need to be great (even 50kb/s would be fine), latency and chaos are more important factors for my use case. I have tried mobile networks for testing but they're too reliable these days, at least where I live. Also tried Starlink once in the woods, and that was unstable enough to work for this use case, but heading out to the middle of nowhere every time I want to ensure that my clientside handling of weird network conditions is... very inconvenient.
I agree that IPv6 is better for testing over, in fact I've already been using it. I use
netnsto trap my client's network access to aveththat rotates to random IPv6 addresses in my /64 during my torture tests to ensure that my application sessions survive IP changesPut a fan in front of the starlink so it partially blocks the signal? I'm only kind of joking.
It wasn't my Starlink equipment (I was on a trip with a friend) and I'm lacking the spare dosh to sign up for a sub of my own. (At this point I'm tempted to start writing my own TCP and UDP chaos proxy for greater entropy because none of the existing chaos monkeys on the market are chaotic enough for my taste & for better lang integration with my favorite languages than Toxiproxy)
Find a friend in mainland China :-)
If you have not tried it yet, mosh is great for ssh on unstable / laggy connections (client side prediction + autoreconnect)
Definitely not mzunguhosting.ml. Newtork is fast my man.