[00:37] *** m0unds has quit IRC (Ping timeout: 265 seconds) [00:38] *** m0unds has joined #arpnetworks [05:13] *** novae has quit IRC (Ping timeout: 264 seconds) [07:16] *** qbit has quit IRC (Quit: leaving) [07:54] *** novae has joined #arpnetworks [07:55] *** qbit has joined #arpnetworks [07:55] *** qbit has quit IRC (Client Quit) [07:56] *** qbit has joined #arpnetworks [07:56] *** qbit is now known as Guest14252 [07:59] *** Guest14252 is now known as qbit [08:42] *** qbit has quit IRC (Quit: leaving) [08:43] *** qbit has joined #arpnetworks [08:43] *** qbit is now known as Guest37399 [08:44] *** Guest37399 is now known as qbit [09:41] *** mkb has quit IRC (Remote host closed the connection) [09:44] *** mkb has joined #arpnetworks [11:14] At first glance it seemed that ARP had engaged in this newfangled Twitter thing with a hashtag on the front page, but I was pleasantly surprised to see that it was actually an IRC channel. [11:14] haha [11:14] irc had # first. [11:15] I think they took it from IRC [11:15] most people don't know what irc are these days [11:40] yeah we own the # [11:51] do the kvm nodes emulate cirrus logic video? [11:52] looks like they do hmm [13:01] (that's what mine says - it's a pretty common default for kvm/qemu) [14:22] *** jcv_ has quit IRC (Quit: leaving) [14:22] *** jcv has joined #arpnetworks [18:44] *** mkb has quit IRC (Ping timeout: 260 seconds) [19:12] *** mkb has joined #arpnetworks [19:21] *** mkb has quit IRC (Read error: Connection reset by peer) [19:28] *** mkb has joined #arpnetworks [19:34] *** mkb has quit IRC (Ping timeout: 245 seconds) [19:40] *** mkb has joined #arpnetworks [19:41] up_the_irons: I'm still getting terrible IPv6 throughput from my native IPv6 connection, while speedtests show 85 Mbps. [19:41] Any thoughts? [19:44] 73 Mbps to a speedtest in San Jose. But 440 Kbps via SCP to matthoran.com. [19:44] mhoran: ipv6 connection from arp? [19:44] I have native IPv6 at home, to my ARP VPS. [19:44] what ISP? [19:44] Time Warner. [19:45] ah. Comcast here, also with native IPv6 [19:45] Nice. [19:46] do you know if it's a congested ISP link? [19:46] ie, packet loss to ARP [19:46] I don't see any packet loss, just low throughput. [19:46] I pretty much have an open SSH connection to my server 24/7 and I never really have a problem with it, just when I try to transfer data over v6. [19:47] do you have a traceroute you'd be willing to share? [19:47] Looks like it takes HE.net most of the way, though Inever had a problem with my HE.net tunnel. [19:48] mhoran: try a udp test maybe? [19:48] hm. I'm pretty sure HE <-> ARP is okay [19:48] *** toeshred has quit IRC (Quit: WeeChat 1.0.1) [19:48] Hmm. [19:49] -6 -O /dev/null http://ec2.acfsys.net/videos/comme_un_enfant.webm [19:49] acf_: https://gist.github.com/mhoran/ef24133f089cfa996d74 [19:49] Gist: "https://gist.github.com/ef24133f089cfa996d74" [19:49] i'll see what my ipv6 thrroughput is like [19:49] wget^ [19:49] that's an HE tunnel to Amazon AWS [19:49] Getting faster and faster. [19:50] Almost 80 Mbps. [19:51] Around the same latency on a traceroute to that EC2 host. [19:51] hmm how do you go to ipv6 address with iperf [19:52] i'm putting it in square brakcets [19:52] and it's not working [19:52] Hmm. I'm not sure. [19:52] i'm doing the -V to set ipv6 too [19:52] meh i'll make a name [19:52] did you escape the square brackets? [19:52] Route to that EC2 host hits the same exit node from Time Warner, so I don't think that's what's overloaded, or where the problem is. [19:53] traceroute back to you: http://paste.unixcube.org/k/9873d0 [19:54] i'm getting 8 to 10 megabit [19:54] well i tried ' and "' [19:54] that's sending from arp [19:54] now i'll test receiving to arp [19:54] 10 megabit send [19:55] i'm hitting he.net in palo alto through to arp it seems [19:56] and ntt back from arp [19:56] mercutio: this is your native ipv6? [19:56] this is native ipv6 at both ends yes. [19:56] openbsd to openbsd [19:58] root@phoenix:/home/acf# wget -6 -O /dev/null http://www.arpnetworks.com/100mb.bin [19:58] 8.08MB/s [19:58] from Comcast [19:58] bah iperf udp mode is binding to a diff address [19:58] Yeah I get < 100KB/s from here. Same as my VPS. [19:59] Compared to wget -4 which reaches 7 MB/s. [19:59] So something on HE.net between NYC and LAX is terrible. But Seattle is fine. [20:00] i'm getting 0% packet loss with udp [20:00] oh i got 1 packet out of 8494 dropped [20:01] i'mg etting shit loads of packet dsropps uploading udp from arp [20:01] hmm [20:02] 80% + packet loss at 4 megabit [20:02] and 0% at 1 megabit [20:02] i have the ipv4 udp limit lifted on my vps [20:02] i dunno if it's separate for ipv6, but even that is at 5 megabit by default i thik... [20:03] there is a limit on IPv6? [20:03] not sure [20:03] are you going via NYC also? [20:04] ohn it may be ip i am using [20:04] willi switch to a diff ip [20:11] yeah 10 megabit is fine on other ip [20:12] not via nyc, via palo alto for me [20:12] oh this address is he.net both directions [20:13] so yeah it may be he.net have difficulties in some locations. [20:14] hmm i have a vm in uk with ipv6 i think [20:14] that should be via new york [20:15] mhoran: wget -6 -O /dev/null http://\[2605:2700:0:5::4713:9574\]/100mb.bin [20:15] another IPv6 endpoint on HE in LAX [20:16] Speedy! [20:16] blodoy hell [20:16] it's vultr and it wants me to use router discovery [20:16] i'll try another one [20:16] Not incrediblyspeedy, but 5 times faster. [20:16] 500Kb/s instead of 100. [20:16] Still not great ... [20:16] how different are the traceroutes? [20:17] my other uk host gets 2.7mb/sec over arp ipv6 [20:17] to that odwnlaod file. [20:17] Quite different actually. [20:18] https://gist.github.com/mhoran/89125af09f10232223a5 [20:18] Gist: "https://gist.github.com/89125af09f10232223a5" [20:18] http://ipv6.uk.meh.net.nz/testfile.zip [20:18] what do you guys get to that? [20:18] from arp i'm only geting 100k/sec [20:18] but that host downloads fast from arp [20:18] Just under 7 MB/s. [20:18] hmm [20:18] Pretty good to that one. [20:20] oh wow :) [20:20] mercutio: does that go to arp via NTT? [20:21] nope [20:21] ok good. return path is NTT.. [20:21] http://pastebin.com/vmNF2SYK [20:22] heh you can see return path is ntt [20:22] from hop 6 :) [20:22] but also, traceroute works with ipv6 automatically on linux now [20:25] how long has this been going on for? [20:27] I only switched to native about 6 months ago but things got unbearably slow maybe a month ago. [20:28] oh :/ [20:28] have we confirmed that it's likely HE? [20:28] their NOC is fantastic... [20:29] Sure seems it. [20:30] I'll send a support ticket tomorrow if up_the_irons doesn't notice this. But it's bed time for me now. [20:42] i really should switch ipv6 on at home again [21:21] *** acf_ has quit IRC (Ping timeout: 256 seconds)