All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Measuring real life VPS connection speed
Hi fellow users.
I have a question, i currently own a KVM VPS, over at HostHatch.com, the node is located in NL. The VPS provider sells this kind of VPS box with 1Gbps connection.
The problem i am facing is i host a few websites, using Nginx, PHP, MySQL stack. All of them are meant to be accessed from Romania. After a few days of setting up the server i found that my main website is loading slowly sometimes, but there is no traffic on the website since its new. After that i made a zip file containing all of my websites on the server to download to my desktop at home, for backup, the file had 800MB. I used CyberDuck to download the file over sftp, the average speed of the download was 800KB/s, i thought something is odd, i went on speedtest.net and made a test to see the download speed i had, and for my surprise it was 100Mbps, thats the speed i am getting form my ISP.
After this i searched the web and found this nice little script . I have used this script to do some manual tests, with servers from Romania. The result was not clear, then i wrote a little script in PHP to run the speed test using the python script i mentioned earlier, but now i am running the script 3 times, on different servers and storing the results in a .csv file. Using a crontab i run this script every hour of the day.
After a few days i downloaded the .csv file and i was amazed! There are times when the upload speed of the server drops below 1Mbps, tested on 3 different servers. The servers are located in the biggest romanian cities, and hosted each by a different ISP (the biggest ISP in the country).
I made a ticket on the support page, and this is the response:
Unfortunately we cannot rely on tests taken from speedtest.net. It is a service made to test home ISPs, not datacenter grade connections.
There are many other more suitable providers that offer test files that you can test from. Examples are OVH, Leaseweb, i3D, Softlayer, etc. Please test with their test files and let me know if you see any unusual speeds and we will be more than glad to check it.
Best Regards,
Mike W
Technical Support @ HostHatch, Inc
I am sharing the results file, for you to analyse.
What is your opinion?
Comments
They were pretty clear with you in the ticket. They can't promise you 1 gbps to Romania and no one can.
If you still want to test the connection speed from the server, check with other ISP in The Netherlands as they suggested, like LeaseWeb or i3D.
Did you see the file with the tests? There are times when the connection is BELOW 1Mbps!!!
Test it using the SAME server each time, some of the servers in speedtest.net are shite. You could also run an iperf server on your home connection and have it connect to that, it would provide a much more accurate result for your own connection.
And? Connection could be congested at any point in the way to Romania, most likely not their fail.
The data you have doesn't conclude anything so far. And you're not the only host on that 1Gbps node. Thus you won't get a guaranteed 1Gbps every second. Your neighbour(s) on the same node might utilize more/less at any particular time too.
@sandbox123 yes, correct! But i think that if i pay, for the server i should have a fair share of the connection. Then they should not advertise that they have 1Gbps connection if, you are not going to see that speed ever. I don't want 1Gbps all the time, but at least to have a 10Mbps upload.
We already said it: test with a nearby network (LeaseWeb, i3D) and you'll see.
I just looked at the ticket and the problem here is that your testing method is speedtest.net. I think anyone will tell you the same, it is not a testing method for servers.
We use 10 Gbps connections (or at least multiple 1 Gbps on older servers) per node and limit each VPS to 1 Gbps, so this isn't the case here. That is why we suggested that he test with those mirrors, which will show that this isn't congestion on our side.
I totally agree with you on the part that you should get what you have paid for. You are getting a fair-share of the connection. Just because you are not able to get a certain amount of speed to a certain ISP in one location on speedtest.net - does not mean it is congestion on our side.
@Abdullah Ok, then speedtest is not the way to test the connection speed, please present a way to determine for certain where the problem is. Since in the file i have provided, there a periods of time where the service works ok-ish ~60-80Mbps, and sometimes the connection speed drops to below 1Mbps! on all of the test servers! As i stated in the post the servers are located in 3 different cities on 3 different ISP, the biggest in the country!
And i have a question, 1Gbps connection to where? NL, EU, US?
@slave2anubis you aren't being very reasonable, multiple people have told you what's wrong and still you don't want to understand.
Your connection in Romania can't get 100 mbps to any part of the world. Want further proof? Try this speed tests:
http://speedtest.tokyo.linode.com/100MB-tokyo.bin
http://hk.edis.at/100MB.test
You're failing to understand that distance affects your maximum speeds as well. Latency plays a big role in how strong your connection is to the testing server. If you were to test it with another server that is close to yours, odds are you'll find better speeds.
Its not about speed, its about consistency!
The point of all this discussion is that my connection speed to my VPS is not constant. So if my test is not good, point out a viable test to see where the problem lies. A test that i can automate and run for a few days.
The farthest you are from the server, the lest constant it will be.
On the server:
wget -O /dev/null http://mirror.nl.leaseweb.net/speedtest/100mb.bin
If you get a decent speed there, the problem is somewhere between Serverius and the ISPs you checked. Being Romania, most likely near your end and not Serverius's, if you ask me. Good luck trying to solve that, because you aren't going to.
I don't think I need to further clarify our point of view on this.
This can be best answered with another question. When you buy a 10/20/100 Mbps internet connection for your home, is it to NL, EU or US?
-deleted-
Could I please have the ticket ID? I would like to look into this for you.
Look, I like RamNode and have services with them, but their network in The Netherlands is no better than Serverius, that's for sure.
@Abdullah Sure, it's #707411.
Oh, I forgot, I also changed the sources from the US Debian mirrors to the NL ones - didn't change much.
Thanks - I will work with you in the ticket and get this resolved.
SpeedTest.net doesn't mean anything for VPSes or Dedicated server. It's just for fun.
Didn't see any ticket response from you yet, so I went ahead and set myself up a server of the same exact specifications as you had, on the same node as your VPS is on. I've performed some download tests to clear that "5 minutes to update" confusion. I am not sure why it took 5 minutes, but it wasn't due to the network. I am happy to perform some upload tests for you if you would like as well.
First of all, here is a basic benchmark from the popular bench.sh script:
Some more tests, specifically from Netherlands' test servers (all test files were 1 GB):
To note again, this is exactly the same package that you had and exactly the same node as your VPS was hosted on.
You have mentioned here that you faced these issues in Amsterdam, yet the whole ticket which I went through is about the server you purchased in Stockholm. You were getting low speeds there and asked us to move you to Amsterdam. I do not see any further complaints in the ticket, or anything relating to Amterdam at all. I am happy to perform some benchmarks for Stockholm if you would like.
I can agree about the Serverius part, but he also appears to have had similar slow speed issues (in Kbps) in our Stockholm location, which is inside the Portlane (AS42708) network. One of the best there is in my opinion.
One last thing I would like to note is - we also provide 10 Gbps VPS. These are specifically used by high bandwidth users (for streaming mostly). There is no way that such slow upload speeds that you are describing would be acceptable to these clients, let alone 250 Kbps.
This should also clear the OP's concerns about the port speed they are getting.
I have a container with you on this network. I can only say good things about the service, and it was dirt cheap.
If you need another point of data use curl-speedtest (EU region).
This script test download and upload speeds.
No provider can guarantee high speed to a country far away.
If you want high speed to your home ISP, get a server located in your country.
@Abdullah I didn't answer yet cauz it was 1 am here in Germany when I wrote that post and went to sleep after it. I'll answer to it later, I am currently at work (yes, I work at Saturdays).
I'll do a little video later that shows, that an update DOES take multiple minutes (without any added sources) and about the up and download speeds. The results might be a bit different than I wrote here, but that's because of the shared connection I guess. I swear, I'm not interested in writing bad things about you if they aren't true (why should I?), it's just my experience. So, hold on, I'll answer to the ticket and link the video here later too.
Just did it:
Testing EU locations
Speedtest from Tallinn, Estonia on a shared 1 Gbps port
Speedtest from Paris, France on a shared 1 Gbps port
Speedtest from Milan, Italy [ generously donated by http://www.prometeus.net ] on a shared 1 Gbps port
Speedtest from Dusseldorf, Germany [ generously donated by http://megavz.com ] on a shared 1 Gbps port
Speedtest from Falkenstein, Germany [ generously donated by http://megavz.com ] on a shared 1 Gbps port
Speedtest from Bucharest, Romania [ generously donated by http://www.prometeus.net ] on a semi-dedicated 1 Gbps port
Looking forward to your tests. The shared connection makes no difference here, as I described above. There is no way that multiple Gbps of traffic was being pushed without us noticing, which is the only way you would have 250 Kbps due to to the shared connection. Unlike many others, we do not use a single 1 Gbps line per node and then share it between all the customers. I think any good host does the same here.
I can see that you've had a bad experience, but please do not make any assumptions that it was due to our hardware or the network. You were getting the same speeds in our Stockholm location, based in the Portlane network, which I think many here will tell you is one of the best networks there is. There is no "just.. Wtf." going on in our network, as you described it. It may be the OS/template/way of testing, but it was not the network, which I think this thread was about.
We would have been happy to investigate and tell you why exactly you are having that problem had you give us a chance to investigate. Because I do not see any complaints in your ticket after you moved to Amsterdam. Only a comment here that calls our network "just.. Wtf."
Then, first learn how the Internet works and then stop demanding stupid proofs when it's obvious (for the rest of the would) that YOU are wrong.
Also, it's frustrating that you ignore other people trying to explain you how transit and peering does work in the Internet. I don't have anything to win trying to convince you, I am just another customer which is very happy with the service. If you still don't believe everything is right, I would happily do some testing for you, just let me know and I'll send you my public key so you can provide me temporary access.
@Abdullah @Nyr
Yea yea, I'm sorry about the bad words about your network. But it's still strange, because I never got such low speeds I already got with you, e.g. With RamNode. And that it's the templates fault - I used Debian 7 x64, CentOS 5.9, 6.5 & 7 and nginx 1.6.2 for the httpspeed tests. However, the video is on its way. Please tell me what I did wrong then..
To clarify your comment, you faced the same slow speeds in our Swedish network as well as in the Netherlands. I think that says a lot about why it is not a network issue, at least not on our side. Adding to that, the tests I posted above.
But yes, I look forward to your tests. Hopefully we can resolve for you wherever the issue is.