Howdy, Stranger!

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

Advertise on LowEndTalk.com
Scaleway bandwidth performance
New on LowEndTalk? Please read our 'Community Rules' by clicking on it in the right menu!

Scaleway bandwidth performance

Issam2204Issam2204 Member
edited October 2015 in General

I'm a bit confused about Scaleway network perfomance.

I prefer to use IPERF to test network performance (instead of WGET). As a client, Scaleway does fairly good with some 500 Mbps/s peaks both to OVH or Online.

The story changes when I use the Scaleway instance as an IPERF server.

Results 1 (SERVER: Scaleway - CLIENT: OVH VPS):

Results 2 (SERVER: Scaleway - CLIENT: Online, Kidéchire):

In the same network (Online --> Online), as expected, Scaleway performs good. Can't say the same when I used OVH.

Now, do you experience the same? If you have the possibility, can you perform some tests so that I can see if this is an isolated case?

This is the command I used:

iperf -c IP_SERVER -t 10 -i 1 -w 500 -P 5 | grep SUM

To launch IPERF as a server, type:

iperf -s

Thanks!

Thanked by 1Janevski

Comments

  • Thank you for your information. I'm looking for ScaleWay reviews.

  • afterSt0rmafterSt0rm Member
    edited October 2015

    Some nodes are located on racks full of seedboxes, so, if you get bad network results just keep creating instances until you get a instance on a rack without so many torrent downloaders.

    National Institute of Science and Technology in Digital Democracy - Brazil

  • I don't have anything with OVH, so did this on my QPS dedi and of course, Scaleway:

    Results 1 (SERVER: QPS - CLIENT: Scaleway):

    Results 2 (SERVER: Scaleway - CLIENT: QPS):

    Also, speedtest.net result (hosted by Online itself) also suggest not so good upload speed:

    Hope this helps.

    Thanked by 1Issam2204

    Nothing for now :)

  • Thanks @emdad.

    Scaleway has decent download, but not so much upload. Yes, the price is cheap, but one expects what's advertised (or at least not 1/4 of it).

  • @Issam2204 said:
    Thanks emdad.

    Scaleway has decent download, but not so much upload. Yes, the price is cheap, but one expects what's advertised (or at least not 1/4 of it).

    Drop a ticket to the support with your complaints and ask for your rights. They should maybe move you for another rack.

    Thanked by 1Issam2204

    National Institute of Science and Technology in Digital Democracy - Brazil

  • @EkaatyLinux said:
    Drop a ticket to the support with your complaints and ask for your rights. They should maybe move you for another rack.

    I will, thanks :)

  • Maybe @OnlineJulien or @bene_online could explain these bad speeds on sw. I do had servers with low speeds like those too.

    National Institute of Science and Technology in Digital Democracy - Brazil

  • Or also @mikmak .

  • Lol, "servers" with 10x better download than upload performance ... riiiight.

    Thanked by 1Junkless
  • UrDNUrDN Member
    edited October 2015

    Please run iperf (preferably iperf3) with 1 thread on 1 minute with 10 seconds interval. Do not force the TCP Window size, the kernel will handle that.

    iperf -c $host -P1 -t 60 -i 10

    Thanked by 1Issam2204

    www.urdn.com.ua - ISP in Ukraine.

  • MarkTurnerMarkTurner Member
    edited October 2015

    Use online's iperf server, I was getting 6Gbps from Atlanta to it the other night.

    Then you can see whether it is a local network issue.

    Thanked by 1Issam2204
  • sinsin Member
    edited October 2015

    Here's a Quadranet (in Dallas) VPS using online.net (I just ran the command from that iperf test thread a few months ago)

    [email protected]:~# iperf -c ping.online.net -i 1 -P 10 | grep SUM
    [SUM] 0.0- 1.0 sec 54.4 MBytes 456 Mbits/sec
    [SUM] 1.0- 2.0 sec 190 MBytes 1.59 Gbits/sec
    [SUM] 2.0- 3.0 sec 224 MBytes 1.88 Gbits/sec
    [SUM] 3.0- 4.0 sec 184 MBytes 1.54 Gbits/sec
    [SUM] 4.0- 5.0 sec 200 MBytes 1.68 Gbits/sec
    [SUM] 5.0- 6.0 sec 210 MBytes 1.76 Gbits/sec
    [SUM] 6.0- 7.0 sec 217 MBytes 1.82 Gbits/sec
    [SUM] 7.0- 8.0 sec 191 MBytes 1.60 Gbits/sec
    [SUM] 8.0- 9.0 sec 168 MBytes 1.41 Gbits/sec
    [SUM] 0.0-10.3 sec 1.76 GBytes 1.47 Gbits/sec

    Ok I tried the command that UrDN suggested instead and used my Dedibox to online.net for this one

    [email protected]:~# iperf -c ping.online.net -P1 -t 60 -i 10

    [ ID] Interval Transfer Bandwidth
    [ 3] 0.0-10.0 sec 847 MBytes 710 Mbits/sec
    [ 3] 10.0-20.0 sec 911 MBytes 764 Mbits/sec
    [ 3] 20.0-30.0 sec 906 MBytes 760 Mbits/sec
    [ 3] 30.0-40.0 sec 894 MBytes 750 Mbits/sec
    [ 3] 40.0-50.0 sec 916 MBytes 768 Mbits/sec
    [ 3] 50.0-60.0 sec 972 MBytes 816 Mbits/sec
    [ 3] 0.0-60.0 sec 5.32 GBytes 761 Mbits/sec

    Thanked by 1Issam2204
  • Issam2204Issam2204 Member
    edited October 2015

    UrDN said: Please run iperf (preferably iperf3) with 1 thread on 1 minute with 10 seconds interval. Do not force the TCP Window size, the kernel will handle that.

    iperf -c $host -P1 -t 60 -i 10

    This is using my scaleway instance as server and another vps (with high download speed) as client:

  • Issam2204 said: This is using my scaleway instance as server and another vps (with high download speed) as client:

    Can you try downloading http://ipv4.rbx.proof.ovh.net/files/10Gio.dat for a few minutes and report the average throughput?

    www.urdn.com.ua - ISP in Ukraine.

  • XIAOSpider97XIAOSpider97 Member, Provider

    This speed....

    Is it worthy to switch my KS-3 to this (with additional storage) for torrenting?

    Pump Cloud has been sold on Sep 1 2018.

  • The bandwidth between two of my scaleway nodes is ~110M bytes/s so 880M bits per sec:

    # Server:
    003:~$ cat /dev/zero | pv | nc -l -p 1776
    ^C16GB 0:00:13 [ 112MB/s] [                          <=>  
    # Client:
    $ nc bit-gen-003 1776 | pv >/dev/null
    1.19GB 0:00:10 [ 112MB/s] [                       <=>    
    

    Scaleway -> a linode in London gets next to nothing:

    # Scaleway server:
    cat /dev/zero | pv | nc -l -p 1776
    ^C64kB 0:00:59 [   0B/s] [  <=>
    

    I don't mind because I'm doing cluster computing but it certainly doesn't look good for those of you running web servers. That said, it looks as if they have an outage at the moment so this probably isn't normal. I haven't been with them for long enough to know what is normal.

  • @bitdivine said:
    for long enough to know what is normal.

    Some racks do have por network performance because of tons of seedboxes running on it, but 30/40 mbps isn't normal of course. If you get network speeds that low you should contact the support. Premium bandwidth, on the other hand, is perfect every time.

    Thanked by 1Janevski

    National Institute of Science and Technology in Digital Democracy - Brazil

  • some of these stats looks strange to me,
    what we would need to help debug and improve this :
    "mtr -rwc 120 a.b.c.d" in both directions (Really, this matters) with both IP src and destinations (so we can reproduce)
    iperf on your scaleway server that shows problem
    same iperf on ping.online.net

    afaict I see no single scaleway chassis filling its uplinks, so the idea that "many seedboxes in one chassis" that would sature something is not the problem here

    thank you all
    Mik

    CTO Scaleway

  • @UrDN said:
    Can you try downloading http://ipv4.rbx.proof.ovh.net/files/10Gio.dat for a few minutes and report the average throughput?

    Sorry if I answer so late:

    Which is pretty good. However, I'm more interested in upload speeds.

Sign In or Register to comment.