OVH New SP-32 and SP-64 2017 with E3-1270v6 4c/8th 3.8/4.2GHz and vRack!

Comments

• K4Y5 Member

  April 2017

  @local said:
  How much money to buy? Where?

  Fucking check OVH.com already.
• FredQc Member

  April 2017 edited April 2017

  I have a weird problem with the server now

  [root@bhs ~]# wget http://proof.ovh.ca/files/100Mio.dat
  --2017-04-24 15:52:57--  http://proof.ovh.ca/files/100Mio.dat
  Resolving proof.ovh.ca (proof.ovh.ca)... 192.99.19.165, 2607:5300:60:44a5::1
  Connecting to proof.ovh.ca (proof.ovh.ca)|192.99.19.165|:80... connected.
  HTTP request sent, awaiting response... 200 OK
  Length: 104857600 (100M) [application/x-ns-proxy-autoconfig]
  Saving to: 100Mio.dat
  
  100%[=======================================================================>] 104,857,600 20.0MB/s   in 4.3s   
  
  2017-04-24 15:53:01 (23.2 MB/s) - ‘100Mio.dat’ saved [104857600/104857600]
  
  [root@bhs ~]# service network restart
  Restarting network (via systemctl):                        [  OK  ]
  [root@bhs ~]# wget http://proof.ovh.ca/files/100Mio.dat
  --2017-04-24 15:53:26--  http://proof.ovh.ca/files/100Mio.dat
  Resolving proof.ovh.ca (proof.ovh.ca)... 192.99.19.165, 2607:5300:60:44a5::1
  Connecting to proof.ovh.ca (proof.ovh.ca)|192.99.19.165|:80... connected.
  HTTP request sent, awaiting response... 200 OK
  Length: 104857600 (100M) [application/x-ns-proxy-autoconfig]
  Saving to: 100Mio.dat
  
  100%[=======================================================================>] 104,857,600  111MB/s   in 0.9s   
  
  2017-04-24 15:53:27 (111 MB/s) - 100Mio.dat saved [104857600/104857600]
  

  Full gigabyte for a few minutes then drop to ~20MB/s

  Have to restart networking to regain 1Gbps

  What could cause this ? This is on a Proxmox KVM VPS running Centos 7 or Ubuntu 17
• WSS Member

  April 2017

  Sure looks like traffic shaping.
• FredQc Member

  April 2017

  @WSS said:
  Sure looks like traffic shaping.

  I should say I cannot reproduce the problem on the node itself.
• teamacc Member

  April 2017

  Possibly VM running out of ram to cache the download and having slow disk access?

  Try saving to /dev/null instead.
• FredQc Member

  April 2017

  @teamacc said:
  Possibly VM running out of ram to cache the download and having slow disk access?

  Try saving to /dev/null instead.

  [root@bhs ~]# free -m
                total        used        free      shared  buff/cache   available
  Mem:           3951         285        2774          24         890        3319
  Swap:          1019           0        1019
  
  [root@bhs ~]# curl -o /dev/null http://proof.ovh.ca/files/100Mio.dat
    % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                   Dload  Upload   Total   Spent    Left  Speed
  100  100M  100  100M    0     0  17.4M      0  0:00:05  0:00:05 --:--:-- 16.3M
  

  [root@bhs ~]# service network restart

  [root@bhs ~]# curl -o /dev/null http://proof.ovh.ca/files/100Mio.dat
    % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                   Dload  Upload   Total   Spent    Left  Speed
  100  100M  100  100M    0     0   110M      0 --:--:-- --:--:-- --:--:--  110M
  
• FredQc Member

  April 2017

  Tested the Intel e1000 drivers

  [root@centos6 ~]#  wget https://x86.ca/bench.sh -O - -o /dev/null|bash
  CPU model :  Intel(R) Xeon(R) CPU E3-1270 v6 @ 3.80GHz
  Number of cores : 2
  CPU frequency :  3791.970 MHz
  Total amount of ram : 995 MB
  Total amount of swap : 511 MB
  System uptime :   1 min,       
  Download speed from CacheFly: 101MB/s 
  Download speed from OVH, Beauharnois, Canada: 111MB/s 
  Download speed from Atlantic.NET, Toronto, Canada: 38.2MB/s 
  Download speed from ColoCrossing, Chicago, USA: 25.9MB/s 
  Download speed from Choopa, Piscataway, USA: 85.1MB/s 
  Download speed from Dacentec, Lenoir, USA: 22.7MB/s 
  Download speed from Atlanta (10Gbps pipe), USA: 9.95MB/s 
  Download speed from Volume Drive, Wilkes Barre, USA: 85.0MB/s 
  Download speed from Atlantic.NET, Orlando, USA: 11.3MB/s 
  Download speed from Incero, Dallas, USA: 37.1MB/s 
  Download speed from Atlantic.NET, San Francisco, USA: 26.8MB/s 
  Download speed from QuadraNET, LA, USA: 30.8MB/s 
  Download speed from OVH, Roubaix, FR: 6.16MB/s 
  Download speed from Redstation, London, UK: 16.6MB/s 
  Download speed from Linode, Frankfurt, DE: 13.7MB/s 
  Download speed from Leaseweb, Haarlem: 19.1MB/s 
  Download speed from Linode, Tokyo, JP: 9.58MB/s 
  

  So it seems that the Virtio drivers are at fault...
• WSS Member

  April 2017

  @FredQc said:
  Tested the Intel e1000 drivers
  So it seems that the Virtio drivers are at fault...

  Yes. The less-overhead not-as-deeply-emulated drivers are obviously to blame.