"Alex.Bligh" <amb@xara.net> wrote
I think that what Alex really should say 'No'. What he saying is that for Xara UK-sourced traffic is split like this: 50% is sent to UK peers at the LINX in London 40% goes to the USA 10% goes to dGIX in Stockholm
Doh! Yes. Must get reading lessons. Missed the word "remaining".
So none of the traffic from Xara goes back out to Xara customers?!
Currently I have no foolproof way to differentiate traffic that goes over the LINX destined for in UK and outside UK. There's some work going on to build a list of UK ASes to this end (amongst others) but it hasn't been done yet.
Quite agreed this is hard stuff, thus my interest :)
Jim is quite right (except it works bidirectionally).
Do you see symmetrical flows in and out of the country or do you see more inbound that outbound? -scott
On Fri, 6 Dec 1996, Scott Huddle wrote:
I think that what Alex really should say 'No'. What he saying is that for Xara UK-sourced traffic is split like this: 50% is sent to UK peers at the LINX in London 40% goes to the USA 10% goes to dGIX in Stockholm ...
Do you see symmetrical flows in and out of the country or do you see more inbound that outbound?
There is a third possibility, of course. In our case we see more outbound traffic towards the USA then inbound traffic from the USA, because our UK customers have Web sites popular in the USA. -- Jim Dixon VBCnet GB Ltd http://www.vbc.net tel +44 117 929 1316 fax +44 117 927 2015
"Alex.Bligh" <amb@xara.net> wrote
I think that what Alex really should say 'No'. What he saying is that for Xara UK-sourced traffic is split like this: 50% is sent to UK peers at the LINX in London 40% goes to the USA 10% goes to dGIX in Stockholm
Doh! Yes. Must get reading lessons. Missed the word "remaining".
So none of the traffic from Xara goes back out to Xara customers?!
Yes of course, esp news. between our news machines and customers. However I've discounted all (OK well most) inter-UK-customer traffic before doing the calculation. The figures were measured through the first UK router in our transit AS which peered at the time of measurement with everyone in the UK and still peers with everyone significant traffic wise within the UK. Traffic between UK customers doesn't touch this router. As we are relatively small market share wise in the UK cf market share of (say) MCI in the US, and as the UK is a small % of the internet as a whole most "inter-customer" most traffic is actually between customers and (say) our news machines, nameservers etc. IE if our UK customer base was 1% of the Internet the most difference this would make is 1%.
Currently I have no foolproof way to differentiate traffic that goes over the LINX destined for in UK and outside UK. There's some work going on to build a list of UK ASes to this end (amongst others) but it hasn't been done yet.
Quite agreed this is hard stuff, thus my interest :)
We're getting there. We are developing tools that break down traffic flow stats into inter-AS flow stats, then in turn break these down into the (commercial) sources of such flows. I'd like if possible to consider "what if" scenarios. We already have some tools in perl that "understand" BGP tables and AS paths - currently the main challenge is grabbing the accounting data as we don't have NetFlow :-(
Jim is quite right (except it works bidirectionally).
Do you see symmetrical flows in and out of the country or do you see more inbound that outbound?
That's currently a hard question as we're in the middle of mucking the backbone about but last time I looked when it was set up in a symmetrical way, the traffic was almost symmetric (volume wise). I'm going to see Jag@bt.net next week for a chat about this BTW - I understand you have both been talking about similar stuff before. Alex Bligh Xara Networks
participants (3)
-
Alex.Bligh -
Jim Dixon -
Scott Huddle