Yayzi vs. BT FTTP Speeds & Routing

Tips and guides
1

There probably aren’t that many people interested in this, but I put together a brief comparison of BT FTTP vs. Yayzi.

For the record, I’m in Cambridge and my BT connection has always had excellent routing/latency, and I appreciate having a low latency connection. Being the biggest ISP in the UK, BT also has excellent peering, so it’s a good comparison to make.

All speed tests were performed with iperf3 with a wired computer, and all latency/routing tests were performed directly from the router.

First, the speeds, not much to say here, Yayzi wins:

BT - 903 mbps down, 111 mbps up
Yayzi - 947 mbps down, 943 mbps up

For routing/latency I was really impressed with Yayzi, it’s very competitive:

8.8.8.8
BT: 5.7 ms
Yayzi: 2.7 ms

1.1.1.1
BT: 6.1 ms
Yayzi: 3.1 ms

AWS Ireland
BT: 16.0 ms
Yayzi: 12.5 ms

AWS Virginia
BT: 90.5 ms
Yayzi: 79.6 ms

bbc.co.uk
BT: 4.4 ms
Yayzi: 2.4 ms

OVH Gravelines (France)
BT: 7.8 ms
Yayzi: 18.7 ms *

OVH Erith (London)
BT: 5.9 ms
Yayzi: 15.9 ms *

*You may notice a discrepancy here with the OVH traces. Connections to OVH take a strange route that end up going via NTT into Italy, before transiting into OVH’s France network, rather than going directly in OVH’s London network like BT does:

  5     9 ms     5 ms     3 ms  xe-3-4-3-0.a03.londen12.uk.bb.gin.ntt.net [128.241.9.234]
  6    11 ms    21 ms     5 ms  ae-0.telecom-italia.londen12.uk.bb.gin.ntt.net [129.250.9.2]
  7    10 ms     9 ms    10 ms  213.144.168.153
  8    16 ms    16 ms    16 ms  par-th2-pb1-nc5.fr.eu [91.121.131.72]

Yayzi have been great for support as well, and their response to my findings about OVH was really good, with them assuring me that they will find a way to fix that route. Hopefully I can update the table soon with some better data for that.

If anyone has specific endpoints that they want tested, I can do that for some time while I still have both connections.

3 Likes
2

Hopefully the OVH routing gets better, as I access various services hosted on OVH.

3

It would appear I am not allowed to edit my original post, but good news is that OVH routing has been improved. The connection now still enters Telecom Italia’s network, but never leaves the UK before transiting over to OVH’s UK network. This has improved things significantly:

OVH Gravelines (France)
BT: 7.8 ms
Yayzi: 9.5 ms
Yayzi (before routing improvements): 18.7ms

OVH Erith (London)
BT: 5.9 ms
Yayzi: 8.8ms
Yayzi (before routing improvements): 15.9ms

The reason that BT is still maginally better is that they have direct peering to OVH, whereas for Yayzi it goes to NTT, then Telecom Italia, then OVH. Still a huge improvement though!

2 Likes
4

Hi… thanks for this information… how did you run these tests using iperf3?

2 Likes
5

The iperf3 tests were against public iperf3 servers for example:

iperf3 -c iperf.as42831.net --parallel 10 -p 5310
1 Like
6

Great improvement, wonder if there’ll be less hops over time. OVH do have a POP in LINX, maybe that’ll get used one day instead for routing.

http://weathermap.ovh.net/#pop_lon

Would be great if Yayzi published a peering list, but I can see why they may not want or be able to do that.

1 Like