Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Kimsufi FR crappy network
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.

Kimsufi FR crappy network

Hello,

For a few weeks now I've been going back and forth with kimsufi support about the quality of their network, most notably the upload speed to higher-ping locations. I tested this from multiple locations world-wide and here are the results, together with a FR-based OVH vps, also limited at 100mbit:

Or, when put on a log-log scale, the issue becomes more apparent:

Their support seems to be unable to pin-point the problem, 3 weeks in.

Multiple people that I've spoken to have reported similar results, although some seem to have proper (OVH VPS quality) speeds.

So my request for this community is this: If you have a kimsufi server, please reboot it into rescue mode (or stop using the network and open port 5001 tcp) and run the following command:

iperf -c proof.ovh.ca -t10 -r

In the case that you are affected, the second part of the test will show it most clearly, and will look something like this:

[  3] local IPREMOVED port 38241 connected with 192.99.19.165 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.3 sec  39.8 MBytes  32.3 Mbits/sec
[  5] local IPREMOVED port 5001 connected with 192.99.19.165 port 57809
[  5]  0.0-10.5 sec  5.38 MBytes  4.28 Mbits/sec

In case the secondary test is under 10mbit/s, you're most likely affected by this (as of yet unknown) bug. If so, please post your results, your datacenter and your rack number (which you can find on the kimsufi control panel)

«1

Comments

  • The figures are amazing :p

    Thanked by 1teamacc
  • rds100rds100 Member
    edited June 2017

    Please consider adjusting these via sysctl:

    net.core.rmem_max=16777216
    net.core.wmem_max=16777216
    net.ipv4.tcp_rmem=4096 87380 16777216
    net.ipv4.tcp_wmem=4096 65536 16777216

    Then retest and see if there is any difference in the results.
    Also you can try changing the tcp congestion algorithm via sysctl to see if it makes any difference.

    Thanked by 1teamacc
  • FredQcFredQc Member

    KS-3C BHS-2

    root@bhs:~# iperf -c proof.ovh.ca -t10 -r
    ------------------------------------------------------------
    Server listening on TCP port 5001
    TCP window size: 85.3 KByte (default)
    ------------------------------------------------------------
    ------------------------------------------------------------
    Client connecting to proof.ovh.ca, TCP port 5001
    TCP window size: 94.5 KByte (default)
    ------------------------------------------------------------
    [  5] local snipped port 56215 connected with 192.99.19.165 port 5001
    [ ID] Interval       Transfer     Bandwidth
    [  5]  0.0-10.0 sec   112 MBytes  94.1 Mbits/sec
    [  4] local snipped port 5001 connected with 192.99.19.165 port 42594
    [  4]  0.0-10.1 sec   113 MBytes  94.1 Mbits/sec
    
    Thanked by 1teamacc
  • @rds100 said:
    Please consider adjusting these via sysctl:

    net.core.rmem_max=16777216
    net.core.wmem_max=16777216
    net.ipv4.tcp_rmem=4096 87380 16777216
    net.ipv4.tcp_wmem=4096 65536 16777216

    Then retest and see if there is any difference in the results.
    Also you can try changing the tcp congestion algorithm via sysctl to see if it makes any difference.

    Already tried these, just tried again to confirm. Doesn't help.

    Also: in rescue mode all settings should be optimized for at least 100mbit right?

  • My KS-1 seeding some torrent

    root@storage:~# iperf -c proof.ovh.ca -t10 -r                                
    ------------------------------------------------------------                 
    Server listening on TCP port 5001                                            
    TCP window size: 85.3 KByte (default)                                        
    ------------------------------------------------------------                 
    ------------------------------------------------------------                 
    Client connecting to proof.ovh.ca, TCP port 5001                             
    TCP window size: 45.0 KByte (default)                                        
    ------------------------------------------------------------                 
    [  5] local 37.187.107.65 port 54277 connected with 192.99.19.165 port 5001  
    [ ID] Interval       Transfer     Bandwidth                                  
    [  5]  0.0-10.1 sec   106 MBytes  88.3 Mbits/sec                             
    [  4] local 37.187.107.65 port 5001 connected with 192.99.19.165 port 40720  
    [  4]  0.0-10.2 sec   106 MBytes  87.4 Mbits/sec
    
    Thanked by 1teamacc
  • rm_rm_ IPv6 Advocate, Veteran

    Yeah got 92/87 Mbit on both of my servers in FR. Do you have this issue on just one server, or many? Which DC (RBX-number?)

    Thanked by 1teamacc
  • What a meme.

    Complaining about not getting his full 100Mbit on services from £4/mo lmao

    Thanked by 3hostdare leonari Pwner
  • @GenjiSwitchPls said:
    What a meme.

    Complaining about not getting his full 100Mbit on services from £4/mo lmao

    Problem is that I am getting my full 100mbit on the 3 eur/month vps, but not on the dedi...

  • @teamacc said:

    @GenjiSwitchPls said:
    What a meme.

    Complaining about not getting his full 100Mbit on services from £4/mo lmao

    Problem is that I am getting my full 100mbit on the 3 eur/month vps, but not on the dedi...

    And? You know a VPS is cheaper to run than the dedi?

  • fLoofLoo Member

    Stop telling him hes only paying amount X for server Y. Hes paying for 100 Mbit, no more, no less. Stop bitching around prices now. He got a problem, stick to that.

  • @fLoo said:
    Stop telling him hes only paying amount X for server Y. Hes paying for 100 Mbit, no more, no less. Stop bitching around prices now. He got a problem, stick to that.

    And? I understand he was sold 100Mbit but you should also have common sense. 100Mbit from £4/mo is just absurd even if OVH use a large amount of peering for their bandwidth.

    Can you point to where it says the 100Mbit is dedicated?

  • @GenjiSwitchPls said:

    @fLoo said:
    Stop telling him hes only paying amount X for server Y. Hes paying for 100 Mbit, no more, no less. Stop bitching around prices now. He got a problem, stick to that.

    And? I understand he was sold 100Mbit but you should also have common sense. 100Mbit from £4/mo is just absurd even if OVH use a large amount of peering for their bandwidth.

    Can you point to where it says the 100Mbit is dedicated?

    I get the 100mbit if i use multiple connections, and the 100mbit is fine to close destinations.

    The lower speeds happen to more remote (higher ping) locations, which is what baffles me.

    This thread is not intended as a ovh/kimsufi bash thread, it is intended to investigate how many other people also suffer from this issue, as it is clear that there are people that are not affected.

  • fLoofLoo Member
    edited June 2017

    @GenjiSwitchPls said:
    Can you point to where it says the 100Mbit is dedicated?

    Sorry buddy, i wont join this kind of conversation in this thread as mentioned before.

    @teamacc said:
    The lower speeds happen to more remote (higher ping) locations, which is what baffles me.

    This thread is not intended as a ovh/kimsufi bash thread, it is intended to investigate how many other people also suffer from this issue, as it is clear that there are people that are not affected.

    TBH this is nothing special. The further the destination, the more peerings/exchanges have to get passed (most of the time). The more POPs you pass, the more "expensive" traffic gets. Also it depends on how you route your traffic.

  • @fLoo said:

    @teamacc said:
    The lower speeds happen to more remote (higher ping) locations, which is what baffles me.

    This thread is not intended as a ovh/kimsufi bash thread, it is intended to investigate how many other people also suffer from this issue, as it is clear that there are people that are not affected.

    TBH this is nothing special. The further the destination, the more peerings/exchanges have to get passed (most of the time). The more POPs you pass, the more "expensive" traffic gets. Also it depends on how you route your traffic.

    Yup, hence I bought a vps from the same company in the same location for comparison, as you can see in the graphs it performs just so much better.

  • saibalsaibal Member

    teamacc said: Yup, hence I bought a vps from the same company in the same location for comparison, as you can see in the graphs it performs just so much better.

    That is probably because:
    1. The host node for your vps has a much better connection than your KS-1.
    2. Most neighbors on your vps' host node are idlers.

  • fLoofLoo Member
    edited June 2017

    Here is an example. I'm running a Zabbix-Satellite on a VPS in SBG. You can see i had problems with it around the 21st to round about the 28th of May (1 week in total). All website-monitors went nuts. So its a luck-or-no-luck game more or less. I'd still prefer a Kimsufi over a VPS, just because of the non-shared-resources.

    Thanked by 1teamacc
  • @saibal said:

    teamacc said: Yup, hence I bought a vps from the same company in the same location for comparison, as you can see in the graphs it performs just so much better.

    That is probably because:
    1. The host node for your vps has a much better connection than your KS-1.
    2. Most neighbors on your vps' host node are idlers.

    It's not a ks-1, it's a ks-3c. Although I got it for the promo price, it should not influence the level of service.

    I get the fact that they will be losing money on this dedi, but that's no excuse. If they didn't want to provide the service at a loss, they shouln't have offered it.

  • saibalsaibal Member

    teamacc said: It's not a ks-1, it's a ks-3c.

    It's still a Kimsufi. If its in the same DC, its the same network whether its a KS1 or a KS3.

    teamacc said: Although I got it for the promo price, it should not influence the level of service.

    Totally agree. I also have a KS-3 at the promo price. But I can max out its network just like I can on my KS1.

    However, as @fLoo mentioned I think its just a transient network congestion. I've had slowness on my Kimsufi's too in the past. But they have always resolved. Hopefully, you are not running anything critical on that box and can wait for the congestion to resolve.

  • IshaqIshaq Member

    KS-1 RBX5:

    [  5] local port 49041 connected with 192.99.19.165 port 5001
    [ ID] Interval       Transfer     Bandwidth
    [  5]  0.0-10.1 sec   106 MBytes  88.6 Mbits/sec
    [  4] local port 5001 connected with 192.99.19.165 port 35186
    [  4]  0.0-10.2 sec   106 MBytes  86.7 Mbits/sec
    
    Thanked by 1teamacc
  • dexterxdexterx Member
    edited June 2017

    RBX4 - Rack: 51D02 Kimsufi KS-3C

    [  5] local 46.105.96.34 port 45543 connected with 192.99.19.165 port 5001
    [ ID] Interval       Transfer     Bandwidth
    [  5]  0.0-10.0 sec   106 MBytes  89.1 Mbits/sec
    [  4] local 46.105.96.34 port 5001 connected with 192.99.19.165 port 43049
    [  4]  0.0-10.3 sec   106 MBytes  86.3 Mbits/sec
    
    Thanked by 1teamacc
  • overloaded switches? is that even possible idk

  • VirtualByteVirtualByte Member
    edited June 2017

    @ethancedrik said:
    overloaded switches? is that even possible idk

    Also switches have limitations, however today most switches have full linespeed capabilities. Also bad switches can cause package loss.

  • @VirtualByte said:

    @ethancedrik said:
    overloaded switches? is that even possible idk

    Also switches have limitations, however today most switches have full linespeed capabilities. Also bad switches can cause package loss.

    Not seeing anything like that on my mtrs though, no matter the destination.

  • BHS1 - Rack: T01B30
    Kimsufi KS-3C - 8G i3-2130 1x2TB Server

    [ 5] local IPREMOVED port 37057 connected with 192.99.19.165 port 5001 [ ID] Interval Transfer Bandwidth [ 5] 0.0-10.0 sec 85.2 MBytes 71.3 Mbits/sec [ 4] local IPREMOVED port 5001 connected with 192.99.19.165 port 33660 [ 4] 0.0-10.1 sec 112 MBytes 93.1 Mbits/sec

  • They apparently just put the upload speed limit on 1gbit on one of my machines, the other seems to be a nice 100mbit.

    Now the download speeds are in the toilet again, resembling the graph in the OP... No word from support about "having changed something, please test again" about it either.

  • sinsin Member
    edited June 2017

    teamacc said: They apparently just put the upload speed limit on 1gbit on one of my machines, the other seems to be a nice 100mbit. Now the download speeds are in the toilet again, resembling the graph in the OP... No word from support about "having changed something, please test again" about it either.

    Wish you luck on getting it fixed, my experiences dealing with Kimsufi and network issues were a nightmare. I moved to using their VPSes instead.

  • rm_rm_ IPv6 Advocate, Veteran

    teamacc said: They apparently just put the upload speed limit on 1gbit on one of my machines

    What. The KS3C is 100 Mbit upload and 100 Mbit download, what 1 Gbit limit?

  • @rm_ said:

    teamacc said: They apparently just put the upload speed limit on 1gbit on one of my machines

    What. The KS3C is 100 Mbit upload and 100 Mbit download, what 1 Gbit limit?

    They should be 100mbit. However, one of mine is uploading at 1gbit to the french iperf server of ovh, and 500mbit to hetzner germany.

  • lurchlurch Member

    So one of you machines has gbit instead of 100mbit? I would keep quiet and enjoy it.

  • @lurch said:
    So one of you machines has gbit instead of 100mbit? I would keep quiet and enjoy it.

    I would prefer some OVH-VPS quality 100mbit link over the crappy gigabit I currently have. Less than 10mbit download from OVH CA, which should never leave their network to begin with.

This discussion has been closed.