New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Vultr is silent giant. They are absolutely amazing. Support, service quality, price. DigitalOcean in compare is crap. Linode... I almost broke my leg trying to use their client area. What else is in compare of those 3?
He even plans emergencies! Amazing provider!
Broke your leg? Dude, whatever the fuck you're doing, you're doing it very wrong.
Well I love you anyway, and if you ever want to talk about it I'm here
I love you more, i did try to hop on the train with you but wasn't picked </3
I highly doubt what OP said. Unless Vultr would GIVEAWAY their openvpn servers for everyone on this thread for free forever?!
Good luck with your budget?
Will you guys ever match Vultr's port speed etc? DO while a good service, still has lots of room for improvement. I use both extensively but Vultr is definitely the beast when it comes to network performance.
thank you for the detailed review
I love it
I can always appreciate you using what works best for you, and wouldn't try to talk you out of it. I can say that our hypervisors are on 40G ports, I'm not sure how that compares. I don't particularly like to compare, I just like to wake up every day and do right by good people until my head hits the pillow again
Vultr VPS are 10 gbit port speed, DO is a mere 1 gbit. I use DO, I just invest more at Vultr for the better network performance / network throughput / etc.
Size matters!
@wdog When you set up openvpn, check the buffer sizes, both client and server, default or too small buffers may kill your achievable tunnel bandwidth, especially when having limited processing power.
While he is very smart, in this thread he's just posting common sense about how networking works... can't argue with the charming part though because that's just how he rolls.
Thanks for the relevant comment, actually addressing what matters to me.
If I left all openvpn installs on all VPS instances with the default settings and some of the VPSs gave me full throughput and others didn't, could I conclude that it wasn't buffer size limiting my transfer speeds on the slower ones? The CPU is never being remotely challenged on any of these instances, touching 10% sometimes.
I don't doubt he's smart, but so far he has only sneered at the thread and most certainly hasn't posted any "common sense". Unless an annoyed "that's not how it works" constitutes wisdom.
There's tons of potential reasons for low throughput.
To find the cause, start by simplifying things. Eliminate OpenVPN settings as a factor by just using iperf to test throughput. On the server, run
iperf -s, and on your computer, runiperf -c [server IP]. Compare that across different providers.If you get high throughput with iperf but low throughput with OpenVPN, then probably you need to tweak some settings.
On the other hand, if you get high throughput on Vultr, but low throughput on other providers, there's still more to test. Next, instead of running
iperf -cfrom your computer, try running it from another VPS. If you get high throughput now, that's an indication that your ISP has a better connection to Vultr. This doesn't mean that Vultr necessarily has a "better network" overall -- maybe they really do have a better mix of upstream ISPs, or maybe it's just your particular ISP and other customers wouldn't have a good connection. (If you're just using it for your own VPN, of course, it doesn't matter since it works for you.)If you get low throughput with iperf for a particular provider from not only your computer but also a couple of other VPS, then probably that provider really is at capacity somewhere in their network. mtr might provide hints about the problem.
There's plenty of further testing you could do from there but that should very quickly give you basic information.
OK, well I'll go ahead and explain it for those not familiar with how networking works.
To get a "good" connection to a server, it's all about location. The quality of a network for a use case like a VPN is definitely subjective and highly dependent on the user's and the server's location, not so much the network speed or server specs. While Vultr may indeed have a good network for you, it's all relative to the routing done between them and the user, something which is not entirely in the provider's control so telling providers who don't perform as well as Vultr that they fail is not fair because it's not really up to the provider how well the routing is between them and you, sure they can take some measures to make it better like better peering and more upstream providers with intelligent routing (such as Noction), but ultimately you and the providers are at the mercy of each hop in between them and you.
Here's the "common sense" part we're talking about, let's assume that everything is 100% equal between Vultr and another provider like DigitalOcean, for the sake of this example they have 100% the same network gear, same upstream providers, same network utilization, and every other minor detail. Now you connect to a VPS in Vultr's network that's in the same city as you, good network performance right? Then you connect to a VPS in DO's network that's not in the same location as Vultr so it takes a different path, the performance will be different because the networks between DO and you are not going to be the same as Vultr and you. It could be better, it could be worse, and some of the hops might me the same eventually but ultimately it's not up to DO or Vultr's control.
So to wrap up this entire explanation with a pretty little bow (and for the TLDR; people):
Networking is unique to each individual person, it's great that you took the time to test other providers until you found one that had good routing to your house, but your line about how other providers fail because their routing is different is not an accurate statement. This also illustrates the importance of test IPs, if you're shopping for a VPN you can save yourself a lot of time, money, and headache by simply grabbing a test IP (open a ticket if one is not publicly available) and running a traceroute to see how many hops and what the latency is between you and them. Aim for under 20ms for a good experience. Speed tests are kind of important, but even if you're maxing out your network port on a test file it doesn't mean anything if you're seeing 100ms latency.
Hopefully that explains things better.
I don't see what the big deal is, setting up openvpn servers on cloud providers. I mean I understand, we've all been there, but I got myself a PIA account for like $2/mo and I love it. Performs beautifully, I get a choice of a slew of servers and bandwidth is unlimited. Vultr (or any other provider for that matter) will not get even close to that. NordVPN costs just about the same if you prepay 3 years and they even have Netflix clean servers. How many Vultr IPs can you watch Netflix on? Zero.
Except that test IPs are not always routed the exact same way as the actual IP of your VPS, and that routes can change anytime.
That is correct, but it's still a better test than say downloading a test file.
Can you point out where he posted common sense and explained how networking works? I just see the comment saying that's not how things work without any elaboration. No one mentioned guaranteed line rates but him.
Regardless, this kind of testing is done all the time (ie, mobile network testing, DNS servers, CDN servers, Antivirus updates, etc). You test large enough data samples and you can say one Datacenter performs better than others. It's statistics. Saying it can't be beat is hyperbole, though.
Then there's Datacenter networking, like 40GB vs 10GB vs 1GB switches, redundancy and over subscription, etc. Datacenter infrastructure is definitely a bottleneck in the system.
So yeah, this is a discussion and I saw nothing useful in Clouvider's post. So maybe I missed one.
Sorry about that, it's common sense for people who understand the basics of networking which I incorrectly assumed that somebody doing a review specifically of a network had that understanding. I corrected this by explaining it in more detail with examples that everybody can understand regardless of their level of networking knowledge.
It can be, but when it comes to latency it rarely is. The routing has more of an impact than the data center's infrastructure because even if the techs don't know what they're doing the chances of them messing it up so badly that you're look at over 5ms of latency before you leave the data center is almost impossible to achieve and would definitely be corrected within the first month of gaining clients.
This won't have much of an impact on latency whereas a dedicated port might. Having a guaranteed speed on a shared port is good for throughput, but not necessarily for latency.
Purely anecdotal, not scientific or anything special. Literally just downloading a few test files at the time of writing. But Vultr is able to saturate my home connection (albeit terrible) both from LA and NY. Something that living in Australia, anything across the ocean usually struggles to do.
I'm actually fairly impressed to be honest.
Where's reviews section/category?
https://www.lowendtalk.com/categories/reviews
It definitely varies. Vultr SG pulling a file from AWS London struggled to get over a few megabit/second. Where as Linode SG and Lightsail SG was able to pull it at ~200Mbit/second. With one transit provider and a peering handoff to equinix they weren't able to do much about it.
tested mainly Tokyo instances for Asia connectivity. Superb. Their $10 advertised 2GB single CPU xeon really works well and fast. Exceeds 3Gbit within Japan (maybe hosted in softlayer?)
Decided it was too much for testing , tried another Tokyo 1GB instance. Still as speedy, installed openvpn access server and using VPN with no issues around 40mbit.
I think their SSD speeds for the new Xeons are also fast, I/O ~ 500MB/s
Of course it's the Lunanode rep who posts the most helpful, on-point info in the entire thread, minus the ego, annoyance and other issues that plague other "knowledgable" posters. Thank you.
Let me say here, as I did in the Lunanode raffle thread, Lunanode is among the absolute best providers. My first VPS, and I still have it and love it and use it all the time (no idling). Totally pro and polished organization with a ton of features. Excellent control panel. Great, attentive support. Privacy and professionalism is a priority for them. They'll even use pgp to encrypt all e-mail communications with you if you upload your public key. Just functioning on a higher level than most providers. I could not recommend any more emphatically.
True, way better response than others. But to be a nag, iperf is buggy and needs more experience than one off uses on different iperf versions on different OS's to know whether iperf is fucking up or the pipe. But without going into a longer rant, iperf tcp testing should also be done with various parallel streams. I suggest 3 to start. Then there's udp testing and packet loss differences, etc. Buffer sizing, etc.
Iperf is used because it is free and available for many operating systems. But anyone who's used it for extensive testing knows how buggy it is (I've opened several bugs over the years). Something to be aware of. Still a tool in the toolbox, but know it's limitations.