#arpnetworks 2015-09-11,Fri

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)

WhoWhatWhen
up_the_ironsSept. 10 was a quiet day... [04:12]
............................................................................................. (idle for 7h44mn)
kellytkIn what context up_the_irons? [11:56]
dacait didn't rain kellytk [11:59]
kellytkdaca, I don't understand [12:03]
.... (idle for 17mn)
mkblike the entire scrollback fits on one page [12:20]
kellytkmkb, got it, thanks [12:21]
.............................. (idle for 2h25mn)
mercutioup_the_irons is the only one who spoke on my friday 11 sep, (and i suspect his thu 10 sep) [14:46]
......................... (idle for 2h3mn)
up_the_ironsmercutio: kellytk : for me, there was no scrollback at all on Sept 10. Just joins / quiets. That's what I meant by quiet. [16:49]
mercutioup_the_irons: but you spoke, and broke the silence! [16:49]
up_the_ironsfor me, it was sept. 11 already [16:50]
mercutiooh
right
[16:50]
............ (idle for 55mn)
mnathani_s/quiets/quits [17:45]
BryceBot<up_the_irons> mercutio: kellytk : for me, there was no scrollback at all on Sept 10. Just joins / quits. That's what I meant by quiet. [17:45]
........ (idle for 38mn)
m0undsyea, i kept thinking my client dropped again because it was so quiet yesterday
had bad luck with 3 different freenode servers taking a dump, this one seems ok
[18:23]
.............. (idle for 1h7mn)
***RandalSchwartz has joined #arpnetworks [19:30]
RandalSchwartzRandalSchwartz waves [19:30]
...... (idle for 25mn)
mercutioyeh freenode has been a bit unstable.
i got booted within the last week
it's a lot more stable than it used to be though
[19:55]
brycecI was on a stable server for 30ish days before the IPv6 troubles knocked me offline :( [19:58]
mercutiobrycec: are you back to using ipv6?
it seems to be pretty stable now.
[20:00]
RandalSchwartzI'm still using v6 for freenode [20:03]
m0undsi am too, just not from my arp vm anymore [20:16]
kellytkEstimates of when IPv6 will be as generally stable as IPv4? I think if that doesn't occur within two years it could become an issue [20:18]
RandalSchwartzwhat's not stable? [20:18]
kellytkI haven't migrated yet so I can only gauge others' experiences [20:19]
m0undsit's network-dependent
my home ipv6 is fine
[20:19]
RandalSchwartzv6 works as long as your router doesn't advertise false v6 connectivity
and sadly, many of them do.
[20:19]
m0undswhat we w ere talking about was some congestion/weirdness that was happening w/arp [20:20]
RandalSchwartzthere's always fine tuning... :)
O
[20:20]
mercutioconnection to he.net was dropping [20:20]
m0undsthere are devices that have flaky/bad ipv6 support, which leads people to have to disable it (glaring at android here) [20:20]
RandalSchwartzI'm always on freenode via v6. I get dropped about once a week.
that's not bad.
[20:21]
kellytkI'll be giving it a while longer yet [20:21]
m0undsi got dropped 3-4 times yesterday, just kept getting servers that were having trouble [20:21]
mercutioi think that there's a smaller number of large issues with internet these days, but more smaller issues
running peering hot seems pretty popular in the US
[20:21]
RandalSchwartzwhat does "peering hot" mean? [20:22]
kellytkAttractive traffic shapes? [20:22]
mercutioit means they have say a 10 gigabit pipe they're trying to route 12 gigabit through
it was the network neutrality / netflix thing
it often is worse in some cities than others
http://arstechnica.com/information-technology/2014/03/level-3-blames-internet-slowdowns-on-isps-refusal-to-upgrade-networks/
stuff like this
[20:22]
BryceBotArs Technica: "Level 3 blames Internet slowdowns on ISPs’ refusal to upgrade networks" [20:24]
m0undsheh
refusal to "upgrade networks"
e.g. refusal to provide peering without $
[20:24]
mercutiowell sometimes they have settlement free peering [20:24]
BryceBotThat's what she said!! [20:24]
mercutiobut there's a refusal to upgrade capacity [20:25]
m0undsright, without providing a good reason or money [20:25]
RandalSchwartzRandalSchwartz just finished an 1125 mile run in western and central oregon in 4 days [20:25]
m0undsi think when you're using tons of traffic like netflix, there should be some sort of agreement to cover cost
nice
[20:25]
RandalSchwartzexhausted
but totally worth it
[20:25]
m0undss/using/pushing [20:26]
BryceBot<m0unds> i think when you're pushing tons of traffic like netflix, there should be some sort of agreement to cover cost [20:26]
RandalSchwartzthere is absolutely nothing like crater lake [20:26]
m0undsi've never been
haven't ever been to the NW
[20:26]
RandalSchwartzI think it rivals the grand canyon, except that you can take it all in at one view
and then there's 100 views like that around the rim
the pure blue clear water... the volcanic rock... all amazing.
[20:26]
mercutiom0unds: what about if you're pulling lots of traffic like comcast?
netflix offer local caches
it seems a sensible way to go in some ways
but now days you can do 10 gigabit over a single fibre.
[20:31]
m0undscomcast operate an eyeball network, content providers want eyeballs for their content
they do transit too, but primarily that's to provide fat pipes for content consumption
[20:34]
mercutiocomcast offer cable tv
and so netflix competes with cable tv
and reduces their total income available
[20:34]
m0undsdirectv and dish network both take more customers from cable than netflix [20:35]
mercutioso if they can increase the cost of other people providing content it maeks it harder for other people to compete [20:35]
m0undsgiving them way too much credit [20:35]
mercutiobe it through lobbying to make it more difficult for other people to receive internet in areas they serve [20:35]
m0undsthey pioneered streaming content to home, but they don't provide anywhere near a direct alternative to catv [20:35]
mercutioor making it more expensive for content providers to provide content
ahh
idk, i think internet should be free :)
more bandwidth, more content
[20:35]
m0undsmeh [20:37]
mercutioand anything that raises total bandwidth available rather than has smart caching etc reduces the cost of bandwidth
which means that it's more free
the 1gig to 10 gig movement is being REALLY slow
[20:37]
m0undsnot on comcast's network :)
since they're already 100gig in most metros and 10gig to headend gear
[20:37]
mercutiohow many people have 10gbe at home though? [20:38]
m0undsnobody, not really necessary imo
just like local loop gig-e, who cares?
oh look, i can get content in my metro at gigabit speeds...which amounts to nearly nothing
[20:38]
mercutiothat's kind of want somep eople said about gig-e 15 years ago
heh
[20:39]
m0undsi can run a speed test to a local speed test server and be happy i have a gig-e pipe and no content provider to deliver at anywhere near that speed [20:39]
mercutionz is being fast with higher speed adoption recently
but international speeds still suck
and there's not much local content
still, 200 megabit seems to be the new sweet spot, if being in a fibre area
and it's weird, given options beetween 100 and 200 nearly everyone seems to go for 200
but now it's basically 30 down or 200 down
there's low demand for in between
[20:39]
brycecmercutio: I'm travelling (at vBSDcon) so not really using my home Internet besides popping online (VPN) from time to time, so I can't really say how IPv6 is working on my VPS, sorry. I'm connected right now to my VPS over IPv6 and it seems fine, but it's a very small test. [20:41]
mercutiothe content providers being able to do > gbe is the current issue
like i can easily get linux packages at > 80 megabytes/sec
[20:41]
m0undsvbsdcon, fun [20:41]
brycecRandalSchwartz: wow, you really ran over a thousand miles? Isn't OR only about 300mi west to east? That's about 4 trips each way across the state. [20:41]
mercutiobut cdn's etc don't really push that fast normally [20:41]
brycecm0unds: so far :) [20:41]
mercutiopartially because i think they're not using ssd's
or otherwise have performance limitations in their caches.
but akamai, netflix etc are using 10gbe ports these days in general afaik
brycec: all good
brycec: it's easier to say something is broken than it's fixed :)
also latency is more important than bandwidth for web browsing generally given decent bandwidth speeds
s/speeds/capacity/
[20:42]
BryceBot<mercutio> also latency is more important than bandwidth for web browsing generally given decent bandwidth capacity [20:44]
brycecmercutio: very true [20:45]
mercutiobut i'm all for gigabit local-loop networks
i don't think people will necessarily get gigabit that often
[20:45]
brycecbreakage is obvious, working is obvious, but "fixed"... [20:45]
mercutiobut as long as networks are managed decently it shouldn't do any harm
i don't think gigabit is enough though :)
[20:45]
m0undsthat's why there are business minded types running isps, haha [20:46]
mercutiohaha
well yeah most new networks are gpon afaik
which kind of limits 24 houses or such to 2.4 gigabit
they use time division multiplexing i think
and encrypt the data
so you get so many slices out of it
dynamic reallocation of those slices, means that it should be possible to do 2 gigabit for one customer
but if everyone goes hard of course it'd have to limit it.
[20:47]
m0undsas long as my isp's network is engineered and managed such that i can use my full pipe and enjoy consistent latency all day, i don't care [20:49]
mercutiowell things shouldn't really be designed so that you can use your full pipe necessarily [20:49]
m0undsm0unds shrugs [20:49]
mercutiojust so that you get a decent chunk of the available bandwidth
and someone else running 100x as many connections as you should get half the bandwidth
[20:49]
m0undswhich factors into network management
yes
[20:50]
mercutioand you get the other half
rather than them get 99% of the bandwidth
it's actually pretty common that running more connectoins gives people more bandwidth
and intelligent queueing etc is kind of necessary to improve that
but you kind of want intelligent queueing everywhere :)
and the only place it really happens is near end users generally
and network neutrality may harm intelligent queueing on interconnections etc.
[20:50]
kellytkFor anyone familiar with Ping plugin of collectd, what values have you observed for `ping_droprate`? I've observed only 0 and 1 but I'd rather not rely on that, and unfortunately the documentation doesn't cover the value [21:01]
RandalSchwartzPing? http://www.amazon.com/gp/product/0448421658/ref=as_li_tl?ie=UTF8&camp=1789&creative=390957&creativeASIN=0448421658&linkCode=as2&tag=stonehengeconsul&linkId=ANEQ2LFY5RLU5ZFB [21:04]
BryceBotAmazon: "The Story about Ping" [21:04]
brycecRandalSchwartz: ++ [21:04]
RandalSchwartzI had that in a blog somewhere, and I had three fans *send me* a copy of that book!
Gotta love the fans.
[21:04]
brycecheh
"Your rabid enthusiasm is appreciated, but slightly misdirected. Consider donating the book to your local library."
[21:07]
RandalSchwartzand yes, it's good to be famous. never gonna give that up. :)
One of them looked like a fairly early edition.
[21:07]
............ (idle for 56mn)
mercutioyou could read the soruce kellytk
s/soruce/source/
[22:04]
BryceBot<mercutio> you could read the source kellytk [22:04]
.... (idle for 19mn)
kellytkmercutio: http://git.verplant.org/?p=collectd.git;a=blob;f=src/ping.c;h=df2f6da6e6bc69158ebad65e8c47615cd10bfcad;hb=HEAD#l653 makes it clear that it's a float value [22:23]

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)