Kitz Forum
Internet => General Internet => Topic started by: Weaver on July 12, 2022, 04:18:37 AM
-
LINX Scotland was set up to help enhance the UK's internet infrastructure with the aim of keeping traffic local. The exchange reduces the dependency on London
If I want to talk to smo.uhi.ac.uk, which is just a few miles down the road here in Skye, as mentioned in various threads some while ago, then I presume that I have to go through BT-land down to London to get to aa.net.uk and then AA routes me all the way back up the whole country, passing through ja.net maybe and finally into uhi.ac.uk’s WAN starting from maybe Perth and then finally up to Skye where we began. Does that sound correct?
I’m unsure about the details because I can’t traceroute it very well - tried www2.smo.uhi.ac.uk because I’m pretty sure that that server is in Skye, not in Inverness or even in someone’s CDN, god knows where. Also, in the traceroute starting from here, I saw various bits of madness: blocking of pings, so I believe, and even a node that my iPad thinks is in Canada - either a bug in the test tools I’m using or else insane routing by someone.
Musing the other day, I was looking at LINX’s website and I see they have an IX in Edinburgh. How much does that help users in Scotland in performance terms who want to talk to other machines in the Highlands, Hebrides or the Northern Isles?
If an ISP the likes of AA were to have a fit of madness and installed a POP at LINX Scotland, would that be the correct move towards improving Scottish AA users’ performance experience in that scenario? In order to make Scottish AA fans truly happy what would be a next step? Get some (more) big names to park CDNs are LINX Scotland as well? Attract Netflix, BBC, Amazon, Google, eBay, whatever - is that how it works? How does it work out financially? Would it just be a great way of someone like AA losing some money? For the marketing dept, would that attract additional Scottish customers because the ISP could be advertised as more Scotland-friendly? And declaring superior performance and lower latency in Scotland (for certain scenarios) than ISPs that have no local POP and rely entirely on London.
In a case where the likes of Zen have POPs [?] in Manchester and London, how does the required Manchester-London link work ? Or an Edinburgh-London link? Who operates such links ? In these cases would it be LINX themselves dealing with it? Sanity check for self: In the first scenario I don’t see why AA would want to install their own London-Edinburgh link, as currently all their traffic is getting brought to London by either BTW or TT. Is inter-LINX sites’ traffic something that their customers/members don’t have to be concerned about?
I freely admit that I don’t know anything about this stuff so please be kind to a fool. :)
-
Yeah, they don't work here either:
traceroute to www2.smo.uhi.ac.uk (206.189.246.215), 30 hops max, 60 byte packets
1 vt1.cor2.lond1.ptn.zen.net.uk (51.148.72.22) 7.570 ms 7.459 ms 7.330 ms
2 lag-9.p1.thn-lon.zen.net.uk (51.148.73.160) 7.191 ms lag-9.p2.thn-lon.zen.net.uk (51.148.73.174) 7.439 ms 7.528 ms
3 lag-2.br1.thn-lon.zen.net.uk (51.148.73.167) 7.239 ms 7.234 ms 7.226 ms
4 195.66.238.25 (195.66.238.25) 7.745 ms 7.740 ms 7.732 ms
5 138.197.251.129 (138.197.251.129) 8.407 ms 8.400 ms 138.197.251.131 (138.197.251.131) 8.273 ms
6 * * *
7 * * *
8 * * *
9 * * *
10 * * *
traceroute to smo.uhi.ac.uk (198.244.229.140), 30 hops max, 60 byte packets
1 vt1.cor2.lond1.ptn.zen.net.uk (51.148.72.22) 7.652 ms 7.637 ms 7.311 ms
2 lag-9.p2.thn-lon.zen.net.uk (51.148.73.174) 7.303 ms lag-9.p1.thn-lon.zen.net.uk (51.148.73.160) 7.311 ms lag-9.p2.thn-lon.zen.net.uk (51.148.73.174) 7.292 ms
3 lag-2.br1.thn-lon.zen.net.uk (51.148.73.167) 7.353 ms 7.280 ms lag-1.br1.thn-lon.zen.net.uk (51.148.73.153) 7.459 ms
4 * * *
5 * * *
6 * * *
7 * * *
8 be101.lon1-eri1-g2-nc5.uk.eu (91.121.215.119) 8.179 ms 8.154 ms 8.135 ms
9 * * *
10 * * *
I'd say the Manchester <> London link works quite poorly to be honest, it almost doubles my ping as it still goes via London for every traceroute I've tried. It seems to be the usual random assignment of Man or Lon depending on what the PPP session feels like. Its why I suspect I'm back on BTW, as TTB seemed to always use London.
That's one way I miss Origin Broadband, at least in their early days. Because Digital Region was Sheffield based and Origins connectivity was via ASK4, I would get completely local routing. I could ping my friend on the same ISP and cabinet with insanely low latency (at least for VDSL).
-
@Alex: So for some reason Zen’s siting of a POP in Manchester seems to create the opposite of an improvement to the performance experience - increased latency - at times ? In contrast though, the Digital region thing which kept traffic local where it made sense was a success. What has Zen done wrong?
I remember reading that B4RN was going to have a physical link direct from their homeland of Westmorland and Lancashire [?] into Edinburgh, presumably that meaning into LINX Scotland. I wonder if that was all about redundancy rather than something somehow aimed at performance enhancement?
I was thinking, if I consider the latency when I’m talking to my next door neighbour here in Skye, what must it be like for people as far north as Orkney or even Shetland? Their route all the way down to London and back even when just talking to a next door neighbour must be insane, and I wonder what the latency will be like in that scenario?
Do we think that in the age when people are getting 900 Mbps FTTP, latency will seem to be more critical, and some FTTP customers might be moaning about latency even though they are now enjoying the benefits of massive throughput, downstream at least, certainly?
-
Perhaps those of us more technically inclined might be frustrated, but the latency is a heck of a lot lower than FTTC for most people so possibly not in general.
I mean you have to consider, I can traceroute to Europe almost as quick as I could to London on VDSL.
traceroute to www.rajce.idnes.cz (185.17.117.177), 30 hops max, 60 byte packets
1 vt1.cor1.lond1.ptn.zen.net.uk (51.148.72.21) 7.653 ms 7.963 ms 7.457 ms
2 lag-8.p1.thn-lon.zen.net.uk (51.148.73.158) 7.485 ms 7.477 ms 7.488 ms
3 lag-1.br1.thn-lon.zen.net.uk (51.148.73.153) 7.443 ms 7.439 ms lag-2.br1.thn-lon.zen.net.uk (51.148.73.167) 7.383 ms
4 ldn-b3-link.ip.twelve99.net (213.248.84.100) 7.553 ms 7.491 ms 7.574 ms
5 ldn-bb1-link.ip.twelve99.net (62.115.120.74) 7.407 ms 7.391 ms 7.528 ms
6 hbg-bb3-link.ip.twelve99.net (80.91.249.11) 20.227 ms 19.501 ms 20.757 ms
7 * prag-b4-link.ip.twelve99.net (62.115.124.27) 33.244 ms 33.239 ms
8 o2czech-svc076227-lag003718.ip.twelve99-cust.net (62.115.58.199) 33.243 ms 33.706 ms 33.440 ms
9 194.228.21.243 (194.228.21.243) 33.693 ms 33.573 ms 33.494 ms
10 90-181-223-142.rco.o2.cz (90.181.223.142) 33.345 ms 32.957 ms 32.774 ms
11 185.17.117.177 (185.17.117.177) 32.105 ms 32.272 ms 31.998 ms
FTTP has shaved between 6-30ms off my latency. My first hop used to be anywhere between around 11ms and 24ms. Which does make me wonder if it DID route via Manchester sometimes but maybe that was invisible on TTB?
VDSL
traceroute to pingbox1.thinkbroadband.com (80.249.99.164) from 82.69.11.30, 64 hops max, 48 byte packets
1 vt1.cor2.lond1.ptn.zen.net.uk (51.148.72.22) 11.570 ms 11.490 ms 11.589 ms
2 lag-9.p2.thn-lon.zen.net.uk (51.148.73.174) 11.860 ms 11.601 ms 12.103 ms
3 lag-2.br1.thn-lon.zen.net.uk (51.148.73.167) 11.772 ms 12.009 ms 11.634 ms
4 netconnex-gw.zen.net.uk (82.71.254.2) 11.625 ms 11.736 ms 11.586 ms
5 po11-13.bdr-rt3.thdo.ncuk.net (80.249.97.22) 11.830 ms 11.775 ms 11.507 ms
6 po4-31.core-rs4.thdo.ncuk.net (80.249.97.85) 12.308 ms 11.760 ms 11.832 ms
7 pingbox1.thinkbroadband.com (80.249.99.164) 11.534 ms 11.428 ms 11.868 msFTTP
traceroute to pingbox1.thinkbroadband.com (80.249.99.164), 30 hops max, 60 byte packets
1 vt1.cor1.lond1.ptn.zen.net.uk (51.148.72.21) 6.878 ms 7.075 ms 7.161 ms
2 lag-8.p2.thn-lon.zen.net.uk (51.148.73.172) 6.746 ms lag-8.p1.thn-lon.zen.net.uk (51.148.73.158) 6.627 ms lag-8.p2.thn-lon.zen.net.uk (51.148.73.172) 6.801 ms
3 lag-2.br1.thn-lon.zen.net.uk (51.148.73.167) 6.902 ms lag-1.br1.thn-lon.zen.net.uk (51.148.73.153) 7.737 ms 7.822 ms
4 netconnex-gw.zen.net.uk (82.71.254.2) 8.018 ms 7.982 ms 7.964 ms
5 ae11-11.edge-rt5.thdo.ncuk.net (80.249.97.21) 7.989 ms 8.062 ms 8.144 ms
6 te1-51-36.core-rs3.thdo.ncuk.net (80.249.97.72) 7.485 ms 7.385 ms 7.210 ms
7 po5-32.core-rs4.thdo.ncuk.net (80.249.97.90) 7.218 ms 6.899 ms 6.973 ms
8 pingbox1.thinkbroadband.com (80.249.99.164) 6.413 ms 6.519 ms 6.341 ms
Also interesting is that looking back, I see my latency HAS been this low before on VDSL yet real-world web browsing was never this fast. I'd say its the speed, but I remember my friends Cable connection WAS this fast at loading pages, so its all rather puzzling.
-
I remember reading that B4RN was going to have a physical link direct from their homeland of Westmorland and Lancashire [?] into Edinburgh, presumably that meaning into LINX Scotland. I wonder if that was all about redundancy rather than something somehow aimed at performance enhancement?
Redundancy.
-
I was thinking, if I consider the latency when I’m talking to my next door neighbour here in Skye, what must it be like for people as far north as Orkney or even Shetland? Their route all the way down to London and back even when just talking to a next door neighbour must be insane, and I wonder what the latency will be like in that scenario?
What I now write is about classic telephony dating from, say, 1985. Back then, your neighbours in Heasta (let's assume the fourth and fifth up from the shoreline) had a standard telephone service from BT. That service would come from the Broadford TE. Broadford TE (just like that of BSE TE, which serves "The Cattery") is a remote concentrator, connected by a number of optical fibres, to its parent DMSU (Digital Main Switching Unit). (To the best of my knowledge, there are only a double-digit number of DMSUs UK-wide for the telephony service.) Number four decides to make a telephone call to number five and so dials the relevant number. The remote concentrator, in the Broadford TE, analyses the digit string sent by number four. It has sufficient intelligence to recognise a telephone number as "one of its own" and so does not establish a link to the DMSU but handles it locally. The ringing voltage is applied to the pair that connects to number five and . . . blah, blah & blah . . . as expected. In other words, the point of cross-connection of the two pairs is performed at the first point of switching commonality.
One advantage of the above is that if Barry Blunder (say a building contractor from a company based in Essex), with his mechanical excavator, severs all the fibres that link the Broadford TE to its parent DMSU then all Broadford local-based calls would still succeed. Obviously other more remote connections would fail. (Including calls to the AutoManual centre that normally services the Broadford TE. So no operator services would be available. Including triple nine calls.)
Hopefully I've correctly summarised the situation. If not then I'll be hearing from 4candles, licquorice or Black Sheep, to name just three members
-
In the early 1990s I did some development work on a new simple transport protocol that could use various RF WAN network architectures. This variety of WAN offerings in those days included the Ericsson Mobitex network, which offered a service similar to SMS. SMS came along later. A Mobitex basestation could pass messages between devices in the same cell even if connectivity from the basestation to the network core was lost, and that cell became an island, similar to Burakkucat’s tale.
-
I think it needs time, the first step of the process is having the point of presence. But initially of course traffic will still need to go to London as thats where the international links feed into and the majority of UK hosted content is actually hosted, and of course where most broadband ISPs locate their network equipment. But "maybe" over time, some other networks will link into Scotland, the likes of AAISP might setup something up there as well and then you end up with improved latency to some local destinations.
I didnt even know there was a Scotland LINX until you made this thread.
https://www.linx.net/about/network/ Is interesting to see there is some content providers setup in Manchester as well.
-
Its just frustrating, we know there is peering all over the country but we still get sent to London before hitting the Internet. Doubly so given I've been on an ISP peered in Sheffield.
The network could be so much more resilient if it hit the Internet more locally with a fallback to go via other routes only if that fails. Of course, the cost of those extra links is likely prohibitive or they would have done it.
-
Indeed, we are so centralised on London for infrastructure.
-
I've often thought about this re terrorist attacks. Especially now with moving to VoIP, they could take out most of the infrastructure just by blowing up telehouse, seems kinda dumb.
-
I don’t agree about the terrorism thing. There are so many sites in london on a giant L2 ring, take a look at the LINX website - I hope I’ve got this right, my apologies if I got the wrong end of the stick. I would hope that MI5 and the rest of govt is on to the terrorist risk though, as you say, and I would hope the defences are very tight. IMHO it would be worth powering a site down as a test to see how resilience really works in reality.
-
Interesting, it does seem they have expanded a lot over the years but everything except Telehouse South are right on top of each other, and thats only a little further away. https://www.google.com/maps/@51.5097943,-0.0024564,827m/data=!3m1!1e3
Taking out that campus would severely cripple the UK Internet with most (all?) ISPs using it as their core interconnect. There really should be more diversity than that, eg why does Zen Manchester still just go via Telehouse London rather than peering with Manchester data centres?
My exchange/headend is not that far from the Sheffield data centres yet I go to straight to London most of the time, sometimes via Manchester with an annoying latency penalty.
Of course this is assuming there aren't hidden hops along the way that have some redundant routes.
-
The LINX London Network consists of two separate high-performance Ethernet switching platforms accessible from 16 data centre locations.
Interxion: A Digital Realty Company, Cloud House West << Isle of Dogs
Interxion: A Digital Realty Company, Sovereign House << Isle of Dogs
Equinix, Harbour Exchange LD8 << Isle of Dogs
Equinix, Powergate, Acton LD9 << Acton, Ealing
Equinix, Slough LD4, LD5, LD6 & LD10 << Slough, Berkshire
Interxion: A Digital Realty Company, Hanbury Street, London City* << Spitalfields / Whitechapel
Telehouse North
Telehouse North 2
Telehouse East
Telehouse West
Telehouse South new
Colt, Powergate << Acton, Ealing
Iron Mountain LON-1, Slough << Slough, Berkshire
Virtus, Hayes << Hayes, Hillingdon
Can't speak for any others but the 2 ISP networks I'm most familiar with both have diverse external connectivity. Losing the immediate Telehouse area wouldn't be great but both have private peering and IP transit elsewhere sufficient to run their networks.
LINX isn't a straight L2 network for the curious. LINX 1 uses VXLAN (https://en.wikipedia.org/wiki/Virtual_Extensible_LAN) so has the resiliency advantages of layer 3 networks.
-
@XGS_is_on welcome to the forum! And thanks for that link, reading the RFC now. Some kind of header fatness record? Big jumbo frames might be nice then, especially here, I wonder ?
-
Thank you.
Jumbo frames are not just nice but required. VXLAN doesn't play well with fragmentation. The network operator needs to have them in place in between the VTEPs / Virtual Tunnel End Points.
I know of a couple of providers whose networks are VXLANs. Their customers ride on the technology right now.
-
Understood. With headers that huge… I was thinking of the real 9k jumbo frames rather than eg 1600 bytes ‘baby’ jumbos - that probably wouldn’t be big enough ? Especially not for the header stacks that contain IPv6 headers?
-
VXLAN adds 50 bytes. 1600 should be just fine though can be set higher if desired.
By the time it leaves the VTEP at the other end it'll come out with the same MTU requirements it entered the VXLAN with.