In reply to a post by 23Prince:

Sorry I simply assumed you would pay for 10Gbps and had an SLA you could enforce.

I'm not clear on this. Who should be paying whom for 10Gb and whom should be getting paid credits for that not being reached given that NCUK own network and servers here?

In reply to a post by jchamier:

In reply to a post by 23Prince:

Sorry I simply assumed you would pay for 10Gbps and had an SLA you could enforce.

Let's assume TBB has 10Gbit/sec to LINX. So traffic to any other UK ISP will be via LINX.

You have completely ruined my fishing expedition with your erudite explanation but thank you I guess

DSL Reports with 6 threads, same as the Think Broadband test, rather than the 24 you get by using the cable option:

A Think Broadband result just now - go by the burst rather than the average as the top % burst is the number reported a lot of the time.

1 <1 ms <1 ms <1 ms router.asus.com [192.168.1.1]
2 10 ms 12 ms 9 ms 10.250.64.1
3 12 ms 12 ms 21 ms pres-core-2a-xe-711-0.network.virginmedia.net [213.104.74.245]
4 * * * Request timed out.
5 * * * Request timed out.
6 * * * Request timed out.
7 * * * Request timed out.
8 24 ms 25 ms 37 ms eislou2-ic-2-ae0-0.network.virginmedia.net [62.254.84.62]
9 25 ms 25 ms 25 ms linx-gw2.thdo.ncuk.net [195.66.236.240]
10 23 ms 32 ms 27 ms po4-31.core-rs2.thdo.ncuk.net [80.249.97.85]
11 24 ms 22 ms 22 ms speedtest7.thinkbroadband.com [80.249.106.133]

Copied from other posters post, but shows that the likely problem is in the Virgin Media network, notice how between hop 3 and 8 some 12 ms is lost, not the sort of latency increase you want across a core network. Its not just a busy router responding slowly to ICMP since later hops are still high too.

Some may recall Virgin Media issues with LINX last year and it was not until things really really broke did anything happen.

TBB with the 8 threads most speedtest servers offer as a minimum.

I think a far better question than trying to blame TBB would be to ask why I can get 480Mb/s on the lowest of the low Amazon AWS on a single thread but require 8 of them to get up to 350Mb/s using VM cable.

In reply to a post by MrSaffron:

Some may recall Virgin Media issues with LINX last year and it was not until things really really broke did anything happen.

Indeed. Not until things broke and I, armed with data from you guys to make the case, ended up speaking with a network engineer there directly and the route was then changed... to go around a congested peering link.

I'm not sure why VM are so bad all the time. My first thought would honestly be some kind of QoS or shaping.

In reply to a post by Ignitionnet:

I think a far better question than trying to blame TBB would be to ask why I can get 480Mb/s on the lowest of the low Amazon AWS on a single thread but require 8 of them to get up to 350Mb/s using VM cable.

<straw man>
VM are pushing for market share and winning 'hearts and minds' by selling speed tiers. Not quality, or capability. This may be a limitation of DOCSIS technology, but more likely a limitation of budgets available.

Other ISPs, without their own network, are more concerned about selling connectivity that performs for the speed your line can do.
</straw man>

Lots of my colleagues would pick a 300 megabit home service with variable latency and lack of single threat throughput instead of a 50/60 megabit service that has constant latency and achieves fast downloads in one thread.

General public versus those whom know a tiny bit about networks.

Edited by jchamier (Wed 11-Apr-18 23:12:47)

I still remember back in the NTL days when someone revealed (dont know if was yourself) NTL had 100mbit links to LINX whilst their competitors had gigabit links, and at the same time NTL was also selling much higher speed connections relative to those competitors, VM is still NTL under the hood, virgin media is only a name provided because branson gets a royalty fee for the brand. So if NTL still work with the same principles (minimal capacity expensive, maximum pushed speed to end users) then its no surprise we seeing links been run hot so often.

(note it may have been gigabit links vs 10gigabit links to competitors, cannot remember which way)

Edited by Chrysalis (Thu 12-Apr-18 09:22:03)

There is a dieselgate moment to be had

According to PeeringDB VM now have 2x 1TB links at LINX. Liberty Global have an additional 200Gb.

https://www.peeringdb.com/net/1412
https://www.peeringdb.com/net/346