↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |
Who | What | When | |
---|---|---|---|
pyvpx | for an extra 500EUR/month one's BCIX port can also extend to AMS-IX
those guys are really gettin' around ;) | [00:26] | |
mercutio | bcix?
oh berlin? i think ams-ix is the biggest exchange in the world? bigger than linx. | [00:35] | |
BryceBot | That's what she said!! | [00:36] | |
mercutio | http://en.wikipedia.org/wiki/List_of_Internet_exchange_points_by_size | [00:36] | |
BryceBot | List of Internet exchange points by size :: This is a list of Internet Exchange Points by size, measured by peak data rate (throughput), with additional data on location, establishment and average throughput. Generally only exchanges with more than ten gigabits per second throughput have been taken into consideration. The numbers in the list represent switched traffic only (no private interconnects) and are rounded to whole gigabits. Take into c | [00:36] | |
mercutio | how do you define size? :)
it looks like de-cix is about the same size | [00:37] | |
BryceBot | That's what she said!! | [00:38] | |
mercutio | and linx isn't as far behind as i thought it was
equnix has mor emembers but lower bandwidth going through it also counting by members is only directly attached there's quite a few people on all of them not directly attached but still advertised. and like it's preferable to be advertised on linx as well as ams-ix if on ams-ix from non europe | [00:38] | |
personally i don't think being on ams-ix would give a huge advantage
probably more benefit from level3 being advertised when it is, as level3's got some pretty good europe connectivity | [00:48] | ||
...................................... (idle for 3h9mn) | |||
up_the_irons | mercutio: no they never sent an rfo actually | [03:57] | |
........... (idle for 53mn) | |||
mercutio | up_the_irons: ahh ok
well it looks like looking glass has been working out ok | [04:50] | |
plett | up_the_irons: You Should be at LINX instead of AMS-IX, then I could peer with you directly :) | [04:51] | |
mercutio | i imagine it's harder to get onto linx. | [04:52] | |
up_the_irons | plett: :) | [04:52] | |
mercutio | but i still think that if being on ams-ix should also be on linx.
which means probably doing niether | [04:52] | |
up_the_irons | haha | [04:52] | |
mercutio | it's easy to see the sense in peering locally. | [04:52] | |
plett | mercutio: Shouldn't be any harder. Both have a set of L2 resellers who will resell connections
https://www.linx.net/join/linxanywhere.html | [04:53] | |
mercutio | but peering remotely gets complicated quick, with traffic taking funky routes. | [04:53] | |
plett | Yeah, remote peering probably doesn't make sense for most networks | [04:53] | |
up_the_irons | it's all going over the internet anyway... ;)
not like there's a secret tunnel from ams to lax that will somehow make things *that* much faster | [04:53] | |
mercutio | plett: basically if you do it, you want to start doing it in like 4 or 5 places. | [04:54] | |
up_the_irons | sure, traceroutes will be shorter, but that probably don't really fool anyone that actually cares | [04:54] | |
mercutio | he.net got really screwed up when new york had flooding.
so having "better" providers can help... thing is that was once, ages ago. and it wasn't down, just extremelly lossy/congested. who are GT? GTT? that's level3 isnt' it? oh, nlayer. | [04:54] | |
plett | mercutio: I work for a .uk based network and we don't have any network outside of the country. We peer at the two london IXPs, LINX and LONAP but no others. We've looked into L2 relaying to get to AMS-IX and DE-CIX but it doesn't make sense from either a commercial or network point of view | [04:57] | |
mercutio | plett: isn't dsl terrible there anyway? | [04:58] | |
plett | We try to make it not horrible :) | [04:58] | |
mercutio | and there's like one cable provdier that you probably have to pay to send stuff to
i have a friend who used to live in UK peak time congestion, adsl1.. | [04:58] | |
plett | Yeah. All the regional cable networks got bought out and now operate under the Virgin Media brand (which is owned by Libery Global, who I believe do some US cable too?) | [04:59] | |
mercutio | it seemed surprising to me. you'd think he lived rural or something. but he lived in cambridge.
and he refused to deal with virgin, because he couldn't get them to stop harassing him about the previous owner of house. and wouldn't deal with him because he wasn't account holder. | [04:59] | |
plett | Hah. You say that. Cambridge is a surprisingly small place, and remarkably out of the way of anywhere people would want to run cables | [05:00] | |
mercutio | oh.
i thought it was a fancy tech town i don't know much about the uk tbh | [05:00] | |
plett | It has a university which does fancy tech stuff, and therefore has many small companies nearby who do tech stuff, but not much in terms of network infrastructure | [05:01] | |
mercutio | it seemed to be like what new zealand was like when we had ATM backhaul.
surprisingly new zealand seems to be getting some gigabit fibre now. but there's very little local content. | [05:03] | |
plett | And yes, all the copper phone lines in the UK are run by British Telecom (BT) which is the company formed in the 1980s when the phone network was privatised. | [05:03] | |
mercutio | he moved to the US, and now has congestion issues again. | [05:04] | |
plett | BT are actually being fairly good at rolling out new technology. 80Mb/s VDSL2 is now available to something like 2/3rds of the population, with the majority of the rest getting 24Mb/s ADSL2+ | [05:04] | |
mercutio | but much faster burst speed.
oh nice. yeah that's not bad. | [05:04] | |
plett | There are still some places which can only get 8Mb ADSL1, but that ADSL1 network has effectively 100% coverage | [05:05] | |
mercutio | i think vdsl is good enough for most people atm tbh
1 megabit upload on adsl kind of sucks for skype video and uploading photos and stuff. | [05:05] | |
plett | They also have a GPON fibre-to-the-premises product which can go up to 300Mb/s. That is technically available anywhere you can get VDSL2, but at an extortionate cost. There are a small number of places which have it instead of VDSL, and it costs normal broadband prices there | [05:06] | |
mercutio | ahh cool. | [05:07] | |
plett | There is also a trial of G.Fast coming soon, which will in theory give 100Mb/s over copper, basically by moving the DSLAM to the top of the pole in the street or under a manhole cover | [05:07] | |
mercutio | i reckon fast copper is the way to go atm tbh
it's just so much cheaper/easier. | [05:08] | |
plett | Only because it's already there | [05:08] | |
mercutio | i mean i know fibre is technically better. but it's expensive to instal.
true. i suppose it's a question of when not if. and whether now is a good time to dig up roads etc. new zealand's moving to fibre. | [05:08] | |
plett | Not really any more expensive than copper. Yes, you need a fancy cleaving/fusing tool to put the connectors on the end, but installation engineers already carry ££££ of test gear in their vans | [05:09] | |
mercutio | with speeds of 30/10 or 100/50
but some people are doing fancy plans now already of like 1000/20 and 200/20 the upload speed sucking. | [05:09] | |
plett | Yeah. Even on the fibre-to-the-prem, it's a maximum of 330/30
VDSL is 80/20 or 40/10 here So yeah, all sucky upload speeds | [05:11] | |
mercutio | the upload speed sucking seems kind of random
i think it's so they can charge more for businesses. it's 100/50 on 4g. | [05:11] | |
plett | There are good technical reasons why it's done like it is
But it does suck for people who do want good upload speeds On copper you have a maximum frequency which the line can cope with, and you split it up so some frequencies are for upload and some for download. Most people want more download, so that gets way more frequencies allocated to it | [05:12] | |
mercutio | yeah but on fibre plans gpon can't handle as much upload
but it can handle 2:1 or something ratios fine. | [05:15] | |
plett | And it's tricky to do different things for different lines because they will crosstalk with each other. So if the exchange is pumping out lots of power on a frequency for one line, it can't also be listening on that frequency on the next line because all it would hear is itself | [05:15] | |
mercutio | yeah i know why it's the way it is on dsl
adsl annex m improves over adsl2+ and vdsl is basically fine. but i mean why have 1000/20 fibre :) | [05:16] | |
plett | GPON is a little bit different. I don't know how they do it in other countries, but here, it's a single fibre at 2.5Gb/s back to the exchange from the GPON node
Each GPON node handles (I think) 48 end user fibre lines, and they are purely passively split and everyone receives everything And they then use TDM to timeslice the upload and download channels so everyone gets a share | [05:16] | |
mercutio | it can balance stuff
so each user gets a fair share | [05:18] | |
plett | Assuming a full GPON node with all 48 lines used, each user would get 52Mb/s to play with (2500/48) | [05:18] | |
mercutio | yeah i don't think they usually run them anywhere near full | [05:19] | |
plett | They don't here, because the takeup levels are very low due to the price | [05:19] | |
mercutio | and i think it's like cable in that it's shared frequencys or whatever, but you can oversubscribe | [05:19] | |
plett | It's only a single frequency in GPON (well, two, one for up and one for down) and then TDM timeslicing so everyone can use it | [05:20] | |
mercutio | it's about twice the cost of naked adsl to get 1000/20 fibre here in areas that support it
i think for wholesale cost. maybe a bit less. it's not outlandish though i think the timeslicing can be dynamic i'm not sure how it works though | [05:20] | |
plett | Yes, it can by dynamic
If it notices that one node always uses its timeslots for sending data when another node doesn't, it can allocate that one more timeslices on the fly And download, every node receives every packet and looks at the headers to see what's for it, so the head end in the exchange can decide who gets more or less bandwidth | [05:21] | |
mercutio | cool. | [05:25] | |
plett | The real problem is backhaul costs for wholesale lines. As it's all run by BT, an ISP buying wholesale access has to either have a presence at every exchange or buy backhaul from BT | [05:25] | |
mercutio | and backhaul costs heaps i imagine | [05:25] | |
plett | A small ISP like us can't afford to have a presence at the many hundreds of exchanges, so has to buy backhaul | [05:26] | |
mercutio | can you do a buying partnership? | [05:26] | |
plett | It ends up costing about £50 per Mb/s on 95%ile billing after all the billing tricks and incentive rebates etc
(which is crazy, when I can buy internet transit for £1 per Mb/s when buying in quanties of 1Gb/s) | [05:27] | |
mercutio | or collective.
wow | [05:27] | |
plett | Yeah | [05:28] | |
mercutio | so you don't want users doing more than like 60gb/month?
is there usage based pricing for users? | [05:28] | |
plett | That depends on the ISP, different companies manage it in different ways
We do usage based charging, yes | [05:28] | |
mercutio | ahh | [05:29] | |
plett | Others do an "unlimited" service and rely on network congestion to limit their users throughput | [05:29] | |
mercutio | how much does it cost to do your own backhaul?
i imagine you can't really allocate each unlimited user a megabit of bandwidth at that kind of costing. i probably do over 1 megabit 95th pernctile at home but it's hard to know actually i might not | [05:29] | |
plett | Not too much for the backhaul itself from exchanges, they are well cabled places after all | [05:30] | |
mercutio | i dunno if i stream movies more than 5% of month or not :) | [05:31] | |
plett | The real costs are in having a presence at the exchange, a bank of DSL linecards and people on call to go and swap kit out | [05:31] | |
mercutio | oh i thought this was for fibre | [05:31] | |
plett | The last time I looked, the average per-user allocation in the UK was something like 100Kb/s | [05:31] | |
mercutio | your own dslams seems silly idea
but having cheaper backhaul sounds smart | [05:31] | |
plett | Yes. If we were starting to PoP exchanges now, we'd ignore the copper completely and just buy fibre | [05:31] | |
mercutio | we used to have 48kbit per user average allocation enforced here
or something like that it went up a bit but it meant there was packet loss in peak hours freuqently. and that seemed to be what my friend had :) | [05:31] | |
plett | Then the only kit we need in the exchange is an ethernet switch to take 1Gb/s LX fibres from BT, and aggregate lots of vlans back to us | [05:32] | |
mercutio | yeah sounds sensible.
but you have to do that per exchange not even per city? | [05:32] | |
plett | Depends | [05:33] | |
mercutio | and an exchange is like 5000 people? | [05:33] | |
plett | For fibre, they've been quite clever, and lots of the small exchanges don't have their own fibre nodes at all, they all backhaul to a regional "mega-exchange" | [05:33] | |
mercutio | oh that's good | [05:34] | |
plett | So it's not every exchange that you need to get to
There's something like 5600 exchanges in the UK. Population density is nowhere near even though, so fibre only covers about 500 exchanges but covers 2/3rd of the population | [05:34] | |
mercutio | ahh ok
still, 500 exchanges is a lot to run fibre to | [05:35] | |
plett | Of those 500 exchanges, you'd probably need a presence at maybe 300 to take fibre everywhere
Yes. There's only 20 people in my whole company. I don't want to have to manage kit in that many places all over the UK :) | [05:35] | |
mercutio | yeah.
so it makes sense to band together for smaller providers. | [05:36] | |
plett | Yes and no. On the retail side, we're all competitors of each other | [05:37] | |
mercutio | yeah. | [05:37] | |
plett | It would be like Ford and Honda sharing a factory to save costs | [05:37] | |
mercutio | but when it's giving money to bt..
not really. it's like ford and honda lobbying together to reduce petrol taxes. | [05:37] | |
plett | True | [05:38] | |
mercutio | because they don't want to invest in making electric cars. | [05:38] | |
plett | Hah | [05:38] | |
mercutio | lots of the smaller isp's here have reasonably friendly relations. | [05:38] | |
plett | I've got to run, I've got to paint a wall | [05:38] | |
mercutio | and like lobbyed for cheaped adsl tail costs and so forth.
ok laters. | [05:39] | |
........................................................... (idle for 4h51mn) | |||
jpalmer | rwhy oh why won't they let me get an IPv6 block, and an AS for my house?! | [10:30] | |
plett | jpalmer: I've got an IPv4 /24 and AS for my house. I haven't bothered applying for the IPv6 yet though. | [10:40] | |
........................ (idle for 1h55mn) | |||
pyvpx | jpalmer: they will! you just have to multihome your house | [12:35] | |
BryceBot | That's what she said!! | [12:35] | |
staticsafe | pyvpx: actually you don't have to multihome anymore to get a /24
new ARIN policy - https://www.arin.net/policy/proposals/2014_13.html | [12:36] | |
pyvpx | true. but a /22 is way cooler ;p | [12:37] | |
.... (idle for 19mn) | |||
hazardous | staticsafe: rly
where can i apply i am using almost a full /24 and i would like to consolidate | [12:56] | |
staticsafe | hazardous: policy is already in effect
you need an ARIN account and a ORG | [12:58] | |
hazardous | how much is it? | [13:08] | |
............ (idle for 55mn) | |||
staticsafe | hazardous: $500 | [14:03] | |
..... (idle for 20mn) | |||
jpalmer | hmm, I'd pay $500, for some IPv6
plett: my understanding is if you have an active IPv4 space through ARIN, they'll give you IPv6 for no charge. | [14:23] | |
plett | jpalmer: I'm in RIPE region. Here, end-user allocations direct from RIPE are EUR50 per year. An IPv4 block, and IPv6 block and an AS number each count as separate resources and are each chargable
And since I already get a /48 of IPv6 space from my ISP and I'm not really multihoming, I don't have a need to get v6 in my own name | [14:33] | |
staticsafe | can new ORGs in RIPE get v4 space anymore? | [14:36] | |
plett | Not end user ORGs, no. If you become an LIR then you get a whopping /22 to provide to your end users. Being an LIR costs something like EUR1300/year | [14:37] | |
staticsafe | ah | [14:37] | |
plett | I got my v4 a couple of years ago before they stopped giving them out | [14:37] | |
staticsafe | im going through the process with ARIN right now, its slow going | [14:38] | |
*** | acf_ has quit IRC (Remote host closed the connection)
acf_ has joined #arpnetworks | [14:40] | |
mercutio | a /48 is heaps for personal use. | [14:52] | |
...................................................... (idle for 4h25mn) | |||
*** | forg0tten has quit IRC (Ping timeout: 272 seconds) | [19:17] | |
................... (idle for 1h34mn) | |||
brycec | Anybody seeing Level3 issues? I can't reach wolfman.devio.us over IPv4, stops shortly into Level3, but others are still connected (presumably different transit) and the IPv6 HE tunnel on that host is still up and working.
| [20:51] | |
...... (idle for 29mn) | |||
mnathani | trace from Toronto times out in Miami he.net land | [21:22] | |
.... (idle for 15mn) | |||
brycec | mnathani: For ipv4? Weird... Thought HE was tunnel traffic and some transit.
This long-running mtr shows some very fucked-up routing since I started it weeks ago http://sprunge.us/XeWj | [21:37] | |
*** | Arenlor has joined #arpnetworks | [21:43] | |
....... (idle for 32mn) | |||
Arenlor | up_the_irons: Any ETA on my newest VPS? | [22:15] | |
mnathani | brycec: can you traceroute out of that host?
if you are still connected via v6 | [22:17] | |
............... (idle for 1h11mn) | |||
*** | Arenlor_s has joined #arpnetworks
Arenlor has quit IRC (Ping timeout: 264 seconds) Arenlor_s is now known as Arenlor forg0tten has joined #arpnetworks | [23:28] |
↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |