Howdy, Stranger!

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


OVH strange bandwidth / speedtest results
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 strange bandwidth / speedtest results

Hello. Can anyone give any ideas about this.

I have a VPS in Limburg, I can speedtest to it from the UK and get nearly the full 100mbps port all day long. Or if 50% of the speed is being used by my customers, I get the remaining 50%.

I have a customer in Spain, who can only get around 2mbps from this server at certain times of day (one would think this points to local / ISP issues etc) however:

he goes to nperf.com, chooses OVH Limburg test server (OK it has 10gbps port) and gets between 5-7mbps.
he tests against my Limburg VPS again and gets 2mbps....

How is he getting such different results to the same datacentre & provider?

Might the VPS have different routing/peering to the OVH test server?

Thank for any ideas which I can use to investigate and hopefully pass info on to OVH.

Comments

  • WilliamWilliam Member
    edited March 2019

    paule3d said: How is he getting such different results to the same datacentre & provider?

    Welcome to ONO. I bet it is ONO cable (Tho Telefonica has similar issues.).

    OVH is unusable to UPC (EU), TPSA (Poland), DTAG/Tmobile (mostly, HR/AT/DE/HU etc.), ONO (ES) and various other networks at congestion/peak (and often always) due to their non-peering stance or simply capacity issues. This is hardly new.

  • First-RootFirst-Root Member, Host Rep

    So your customer should go to his (home) ISP and tell them to fuck off and get their job done as they don't deliver what he pays for.

    Thanked by 2poisson uxtvdl
  • dfroedfroe Member, Host Rep

    @FR_Michael said:
    So your customer should go to his (home) ISP and tell them to fuck off and get their job done as they don't deliver what he pays for.

    In depends on the definition of "what he pays for".
    I'd say most German ISPs for example have a clause within their ToS that they can only guarantee a certain bandwidth within their own network and cannot guarantee anything at all (no bandwidth, latency, packet loss) for destinations outside their own network. Which in the end means you have no clue what throughput you will see in real life.
    Maybe the situation is a little bit better in other countries.

  • No, this is not the reality but yes complaining - aside that it will not help - is a way to go forward. Guarantees, if any, are only provided to the first hop after the ISP network; which in most cases is the upstream router in your capital/city.

    Tcom guarantees me 700/700Mbit to Zagreb which i get at any time, but if i cross external i obviously enter a way more shared medium. This is just logical and the speeds are clearly advertised as "up to"; the time of this lawsuits is long over in EU.

    OVH will peer with anyone but eg. as Telefonica with core in Madrid i have no reason to rent/use a wave to Paris to do so - ultimately OVH wants to dump egress to me so at the very least they can peer in my datacenter/city on their cost.

    This is one issue side; the other is that DTAG simply does not peer as it strives (since... 199x...) becoming a Tier1 - Same for Telefonica in a way. Any ISP with any non settlement free peerings (eg. to OVH) does not qualify as Tier1 entirely.

    UPC peers very strictly and got better with OVH but yes, they have backbone issues BECAUSE they agreed to pick up from POPs not best connected to their backbone (among other issues).

    ONO just sucks aside that it does not usually peer; they have a huge client base so they sell you "paid peering" as DTAG does.

  • Hi.

    Thanks a lot for the replies.

    Not sure if I was unclear on my OP, that my VPS is OVH in Limburg, where the client in Spain speedtests 2mbps (I get <100mbps so we know the server is good) yet when he speedtests official OVH server he gets <7mbps.

    Trying to work out why or how this could be, it doesn't make sense, same datacentre and provider, much different results.

    Regards

  • TheLinuxBugTheLinuxBug Member
    edited March 2019

    They have different tiers of bandwidth and also have limitations per rack / per server as these are shared services. The testing server you are trying probably sits on it's own 10Gbit uplink and is meant to be exposed in a way to promote their network, so on their premium bandwidth tier and not sharing an uplink. The VPS and dedicated servers, especially at the lower end, have a higher subscription rate per link than the test server. So this could be as simple as the Hypervisor node your are on has a lot of demand from the server on it at that time, or it could even be the rack your located in has a higher saturation. You can always ticket OVH directly and see if there may be something on their side affecting your speeds and see if they have any suggestions or can maybe migrate you to another node to see if it helps. Outside that, it could be numerous things, IO use on the VPS being high and causing a slow down in reading the file, CPU usage being high and limiting IO availability or network availability, variance in latency,etc, etc.

    TL;DR;
    There are really too many variables to speculate on and ticketing OVH may be you best bet to see if they see anything from their side. Otherwise as William said, over-subscription on the link between you and OVH is probably a large contributor.

    my 2 cents.

    Cheers!

    Thanked by 1datanoise
  • datanoisedatanoise Member
    edited March 2019

    Be it 2mbps or 5-7mbps it's too slow, you can open a ticket with a traceroute and see if they can do something.

  • user54321user54321 Member
    edited March 2019

    can't complain about ovh, atleast to the german upc brand they deliver fullspeed
    and their nperf server in limburg has broken ipv6

  • user54321 said: can't complain about ovh, atleast to the german upc brand they deliver fullspeed

    I can't complain as well

    William said: OVH is unusable to UPC (EU), .. DTAG/Tmobile (mostly, HR/AT/DE/HU etc.).. at congestion/peak

    It's now peak/rush-hour imho

  • Thanks again for replies.

    @datanoise the customer only gets 8mbps maximum, I forgot to say that initially.

    So he gets near max speed to Limburg OVH test server,
    but only gets 2mbps to my Limburg OVH VPS. (which has very little load and will easily do 90mbps to me at the same time the customer gets 2mbps, load is not an issue).

    The customer's ISP is Dinfotec who use Cogentco Spain. They only have private peering (unusual and lame?) and I've found many articles on the net about Cogent having disputes with major Tier 1 providers (Level3, Telia, Verizon etc). https://www.peeringdb.com/net/267

    Both OVH and Cogent use private peering facilities in "Interxion Madrid (MAD1, MAD2)" so perhaps the signal goes through there, but is private peering more likely to be throttled?

  • RazzaRazza Member

    @twiigl said:

    user54321 said: can't complain about ovh, atleast to the german upc brand they deliver fullspeed

    I can't complain as well

    William said: OVH is unusable to UPC (EU), .. DTAG/Tmobile (mostly, HR/AT/DE/HU etc.).. at congestion/peak

    It's now peak/rush-hour imho

    OVH seem to have private peering with DTAG at Frankfurt if you look at OVH weathermap, OVH doesn't seem to be as bad William is making it sound.

  • paule3d said: The customer's ISP is Dinfotec who use Cogentco Spain. They only have private peering (unusual and lame?) and I've found many articles on the net about Cogent having disputes with major Tier 1 providers (Level3, Telia, Verizon etc). https://www.peeringdb.com/net/267

    Both OVH and Cogent use private peering facilities in "Interxion Madrid (MAD1, MAD2)" so perhaps the signal goes through there, but is private peering more likely to be throttled?

    I will probably get browbeat by someone who isn't as opinionated about Cogent, but in general, their routing is 'Okay' but throughput during peak hours is usually not that great, so what your reporting, to some extent, doesn't sound so unusual. Sorry to say.

    However, again, the best way to really find out exactly whats going on is to report it to one or both providers and see if that can explain why there is a slowdown.

    my 2 cents.

    Cheers!

  • nperf uses more then just 1 connection so to saturate a bad link is easier

  • twiigl said: It's now peak/rush-hour imho

    In simple terms without going into the complexity of their and the DTAG backbone: pipe capacity is less issue than per thread speed and especially PP/S.

    Razza said: OVH seem to have private peering with DTAG at Frankfurt if you look at OVH weathermap, OVH doesn't seem to be as bad William is making it sound.

    sure do, at not enough capacity and not all routes - DTAG also covers up to edge of Eastern EU so a FRA only peering is for eg. OVH PL just nowhere near ideal.

    I am DTAG customer myself with one of the (on paper) fastest connections you can get on their backbone (Gbit/Gbit, FTTH) and same as to Hetzner my per-thread caps out at non peak 100-150Mbit and peak at 30-75Mbit to OVH. DTAG internal i can pull 700+ consistently... at a price.

    I sometimes get capped on download because i can not send TCP-ACKs fast enough as well... aka my upload towards DTAG is capped/congested so the empty link you see on LG/map is empty because no one can actually use it...

    My route also by their choice ends up going to Hungary and then back to DE while usually i get either Italy-DE or Slovenia-AT-DE.

    LGI/UPC and DTAG are just into such crap, seen enough in my AT years and UPC...

Sign In or Register to comment.