Howdy, Stranger!

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


OVH packet loss
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.

OVH packet loss

dnwkdnwk Member
edited February 2013 in Providers

I just got a $39 server from OVH(Canada DC). It looks like there is a big packet loss inside their DC, when traceroute

mtl-1-6k.qc.ca (178.32.135.45) 37.169 ms * *
bhs-g1-6k.qc.ca (198.27.73.1) 39.610 ms * 41.552 ms
* bhs-2-6k.qc.ca (198.27.73.16) 251.651 ms 217.392 ms

Comments

  • Package loss hehe

  • jarjar Patron Provider, Top Host, Veteran

    image

  • I think I am going to need popcorn for this xD

  • what so funny?.......

  • HalfEatenPieHalfEatenPie Veteran
    edited February 2013

    You mean Packet Loss.

    Package = Weiners = Ding-Dongs = Fireman

  • jarjar Patron Provider, Top Host, Veteran

    @dnwk said: what so funny?.......

    Experiencing a little loss of your package on valentine's day?

    Maybe it's cultural humor ;)

  • ...........I got it....

  • @dnwk said: I just got a $39 server from OVH(Canada DC). It looks like there is a big package loss inside their DC, when traceroute

    Have you opened a ticket with OVH?

    What is the point of this thread? Nobody here works at OVH that I know of, We can't just walk in there and fix the router for them if that is the problem.

  • @luma said: What is the point of this thread? Nobody here works at OVH that I know of, We can't just walk in there and fix the router for them if that is the problem.

    I know, just let other know the problem. I ask OVH about this. They said their router set traceroute at low priority. Maybe it means their router are too busy.

  • @dnwk what loss do you get with direct ping to outside, ie:

    ping 8.8.8.8

    If you get 0% packet loss, you don't have to worry about any perceived loss within their network. As they mentioned they deprioritize icmp, their routers are busy forwarding packets. In most cases this means there is no impact to end-end ping or throughput.

  • If a routing system is set to put low priority on a packet, and it can't keep up with that priority level, wouldn't that mean the network is overloaded?

    But that's not possible. OVH are the best hosting company on the planet. You should accept the packet loss as something unique and special, as something that signifies a quality host, because they're the cheapest.

  • dnwkdnwk Member
    edited February 2013

    @Damian said: If a routing system is set to put low priority on a packet, and it can't keep up with that priority level, wouldn't that mean the network is overloaded?

    I have the same worry.

    @Damian said: But that's not possible.

    Impossible is nothing.

  • WebProjectWebProject Host Rep, Veteran

    looking on the OVH statuses, they experience the issue with routers:
    http://status.ovh.co.uk/?do=details&id=4111

  • @Damian said: If a routing system is set to put low priority on a packet, and it can't keep up with that priority level, wouldn't that mean the network is overloaded?

    Not really. It could just be a problem with overloaded CPU of the router.
    Usually real traffic is is forwarded in hardware. ICMP time exceeded in transit (i.e. the packets you receive back to make the traceroute work) are generated by the control plane (CPU).

  • @Damian said: If a routing system is set to put low priority on a packet, and it can't keep up with that priority level, wouldn't that mean the network is overloaded?

    @dnwk said: I have the same worry.

    Not at all. As @rds100 points out, routers are built to forward packets, not respond to them directly. Compared to end-end connectivity/throughput packet loss on a router means nothing. (control plane priority, acls, whatever.) as OVH already mentioned they don't prioritize ICMP unless you have actual packet loss to your destination I wouldn't worry.

    Yes, OVH are cheap/suck in many ways but they don't buy small routers or pipes.

  • Check for yourself weathermap.ovh.net/usa

  • @unused said: as OVH already mentioned they don't prioritize ICMP unless

    I also tried UDP traceroute

  • rm_rm_ IPv6 Advocate, Veteran
    edited February 2013

    @luma said: What is the point of this thread?

    To get the actual clueful response like this:

    @unused said: Compared to end-end connectivity/throughput packet loss on a router means nothing. (control plane priority, acls, whatever.) as OVH already mentioned they don't prioritize ICMP unless you have actual packet loss to your destination I wouldn't worry.

     Host                                Loss%   Snt   Last   Avg  Best  Wrst StDev
     1. 192.168.0.250                     0.0%    75    0.3   0.3   0.2   0.9   0.1
     2. 10.92.127.126                    22.7%    75    1.7  32.1   1.2 528.0  96.0
     3. tmn01.transtelecom.net            0.0%    75    1.8  11.4   1.1 177.5  31.9
     4. decix2.routers.ovh.net           98.6%    75   83.2  83.2  83.2  83.2   0.0
     5. rbx-g1-a9.fr.eu                   0.0%    75   85.5 101.5  85.1 434.8  63.2
     6. ldn-1-6.uk.eu                    74.3%    75   94.7 100.0  87.9 155.2  18.4
        ldn-1-6.uk.eu
        rbx-s5-6k.fr.eu
        ldn-1-6.uk.eu
        ldn-1-6.uk.eu
        ldn-1-6.uk.eu
     7. nwk-1-6k.nj.us                   66.2%    75  156.8 159.6 156.3 190.3   7.4
     8. mtl-1-6k.qc.ca                    1.3%    75  164.1 169.4 163.4 352.6  25.6
     9. bhs-g2-6k.qc.ca                  37.8%    75  167.4 167.3 165.2 204.1   6.4
    10. bhs-1-6k.qc.ca                   12.2%    75  224.2 183.2 167.7 436.3  39.4
    11. *my server*                       0.0%    75  173.8 174.6 173.4 192.7   2.8
    

    ZOMG packet loss ...right?
    Nope, zero percent loss to the actual server, and that's the only thing that matters.
    The routers might be ratelimiting ICMP/UDP trace responses, e.g. to prevent overloading/DDoS, like pointed out above such things are handled by the weak CPU, not in ASICs.

    Also appalling to see the amount of OVH-hate and jeering even among those whom I considered to be "respectable providers" in this thread. OVH did nothing wrong in this particular case, but you readily jumped at the chance to post your viagra pic / popcorn / hehe / lfmao / sarcastic statement at them. Don't you feel kind of stupid now? :)

  • dnwkdnwk Member
    edited February 2013

    @rm_ One question. Is it possible that when I ping an IP, the latency is different than actually latency (e.g. when using SIP Protocol), if as you said packet lose is because manual restriction on routers.

    The reason I ask is, I run a VOIP server. Latency is important when you running a VOIP Call. However, recently I move my server and find out that ping to my SIP provider's IP got a significant higher latency than before(200ms difference). However, while I expect call quality to be worse than before, it actually improved. So I am wondering, maybe the latency show up in ping does not always reflect the actual line quality.

  • MaouniqueMaounique Host Rep, Veteran

    Well, carriers do not mangle priority, that happens at the ends.
    ICMP is different for reasons detailed above and may not give a good idea in some cases.
    If the isp is specialized in VoIP it will go to great lenghts to make the network as fast as possible for those connections (there are quite a few possible tweaks in that field) so it is very possible to have a better service with a more distant provider.

  • @dnwk said: it actually improved

    Did any of the other variables affecting call quality changed? Perhaps a low bitrate codec is now being used due to high latency?

  • MaouniqueMaounique Host Rep, Veteran
    edited February 2013

    @MiguelQ said: Perhaps a low bitrate codec is now being used due to high latency?

    High latency does not always mean lower bitrate. Using a lower bitrate codec can help in case of line saturation which would increase latency, of course, even lead to packet loss. But latency usually comes from longer distances or weird routing.

Sign In or Register to comment.