who is aorta.net

so my friend rents a battlefield 4 server in the uk, i believe its in london, my bt connection gives me 11ms, my virgin media gives me 46ms... thats more than the same connection gives me to germany. looking at traceroutes, it seems bt goes directly to the server, whereas the virgin connection is going through aorta.net, this is adding more than double the latency, why is it going through aorta, its the same with my other mates on virgin, all other providers go directly to the server.

link to tracert

https://gyazo.com/9d5c0303d150936e3748d7f277c3d8c6

this it from bt to the same server, the high latency at the bottom is due to a dodgy pcie wifi card, rest of traceroute gives solid low ping.

https://gyazo.com/00ab0944b3d75531ade847966ac455a4

It is just a different route as technically none of them go straight to the server unless you were on the same network and the routes will be different based on whatever network you're on.

Mine for example goes via Linx.

traceroute to 46.251.234.52 (46.251.234.52), 64 hops max, 72 byte packets
1 172.20.20.254 1.010 ms 0.536 ms 0.586 ms
2 193.105.150.125 1.141 ms 1.833 ms 1.383 ms
3 31.13.6.253 3.928 ms 3.915 ms 3.166 ms
4 78.109.176.89 2.203 ms 1.642 ms 1.446 ms
5 81.23.63.228 12.019 ms 5.735 ms 5.581 ms
6 195.66.224.107 5.948 ms 6.023 ms 5.537 ms
7 46.251.234.52 7.046 ms 6.617 ms 6.888 ms

Virgin seem to prefer aorta which is using Liberty Global addressing (who own Virgin, or part of). I suspect they will prefer routes provided by their group and Liberty Global do have peers with Link11 who own the network you're trying to get to.

Routers don't decide paths based on latency, usually automatic but each may be given a preference in terms of cost, i.e preference on peers than transit.

Matt

Edited by uno (Fri 19-Aug-16 17:51:27)

http://www.prnewswire.co.uk/news-releases/chello-bro...

Sky has a low amount of hops to that server.

Tracing route to 46.251.234.52 over a maximum of 30 hops

1 <1 ms <1 ms <1 ms SkyRouter.Home [192.168.0.1]
2 * * * Request timed out.
3 11 ms 10 ms 12 ms be381.pr2.hobir.isp.sky.com [2.120.8.150]
4 10 ms 10 ms 10 ms linx.link11.de [195.66.224.107]
5 10 ms 10 ms 11 ms 46.251.234.52

ICMP hops are at best a vague proxy measurement of how the route is actually constructed no?

In reply to a post by GeeTee:

ICMP hops are at best a vague proxy measurement of how the route is actually constructed no?

You cannot get an accurate idea of routing from ICMP or anything else user visible, as hops inside an MPLS network, inside a tunnelling protocol such as L2TP or between Level 2 switches will not show up.

The first visible hop beyond my router on my Zen FTTC is a Zen gateway in London. This reveals nothing of the cabinet DSLAM, the one or more devices on Openreach GEA between that DSLAM and the fibre Ethernet port where the traffic is handed over to Zen at the exchange, or how the Zen backhaul gets the traffic from that port to the gateway in London.


ICMP can tell you how traffic moves between networks (ASes) using IP routing protocols such as BGP. It is not necessarily a reliable guide as to how traffic moves within those networks.

All that said you do know that any routing done at the not visible layer is not costly in terms of latency, since a hidden route from London to Lisbon and back around be noticeable in the higher latency, even if the only visible hops were Lincoln to London.

In reply to a post by MrSaffron:

All that said you do know that any routing done at the not visible layer is not costly in terms of latency, since a hidden route from London to Lisbon and back around be noticeable in the higher latency, even if the only visible hops were Lincoln to London.

Indeed - any traceroute, pathping, mtr or similar gives an overall round trip time to the furthest reachable point and to all visible hops. This inevitably includes the latency of invisible hops, as will as the effect of DSL interleaving and similar.

The mistake is to claim a particular route is better / more direct / more robust because there are fewer visible hops or because fewer ASes are crossed. You cannot justify any such claim without far deeper knowledge of the network topology and routing than ICMP reveals.

Are you saying that a "traceroute" is a waste of time?