[01:20] http://arstechnica.com/information-technology/2014/11/netflix-takes-up-9-5-of-upstream-traffic-on-the-north-american-internet/ [01:20] Ars Technica: "Netflix takes up 9.5% of upstream traffic on the North American Internet" [01:21] i wonder why it's acking so much [01:22] I read that as 9.5% of traffic leaving NA is Netflix, i.e. foreign users streaming Netflix over VPN [01:22] wow [01:22] netflix is meant to be coming to new zealand in march [01:22] (that's just my interpretation of the headline though) [01:22] oh nah [01:22] this is about uploading [01:23] tcp/ip acks way too much [01:23] i wonder if things like ps4 don't do sack or something stupid [01:25] hmm it may just be that people aren't uploading much [01:25] and it's not a percentage of the total, but of upload in general. [01:25] Indeed, that could just be 10% of 1mbps or 10% of 1000gbps, it doesn't say [01:26] well skype is uploading more than downloading. [01:26] i think it's meant to be of the bandwidth the isp is doing in general. [01:26] But skype calls are far shorter and less ubiquitous than netflix [01:26] but they also use just as much downstream as upstream [01:27] for instance, 4/5 users watch netflix for hours on end, while 1/5 makes a 20 minute skype call [01:27] yeah i was more meaning you can't compare percent upload vs percent download [01:27] because skype should show in downstream table if that's the case [01:27] because it's about 1;1 upload/download [01:28] i dont' watch netflix, and talk on skype for more than 20 minutes sometimes. [01:28] but i know what you mean [01:28] skype is pretty convenient really [01:30] there was this idea years ago of doing ack pacing to remove redundant acks on cable networks [01:30] because upload was congesting a lot. [01:30] and really the whole ack system could do with an update [01:31] but no-one wants to break the internet [01:32] it would especially help 3g/4g/wireless which are half-duplex though [01:36] QUIC may actually help [01:45] *** dwarren has quit IRC (Quit: leaving) [01:45] *** dwarren has joined #arpnetworks [01:46] *** dwarren has quit IRC (Client Quit) [01:47] *** dwarren has joined #arpnetworks [01:47] *** dwarren has quit IRC (Client Quit) [01:48] *** dwarren has joined #arpnetworks [01:48] *** dwarren has quit IRC (Client Quit) [01:51] *** dwarren has joined #arpnetworks [01:55] *** dwarren has quit IRC (Client Quit) [01:59] *** dwarren has joined #arpnetworks [02:54] *** dj_goku has quit IRC (Ping timeout: 240 seconds) [03:07] *** NiTeMaRe has quit IRC (*.net *.split) [03:07] *** up_the_irons has quit IRC (*.net *.split) [03:07] *** toddf has quit IRC (*.net *.split) [03:07] *** twobithacker has quit IRC (*.net *.split) [03:07] *** RandalSchwartz has quit IRC (*.net *.split) [03:07] *** brycec has quit IRC (*.net *.split) [03:07] *** tabthorpe has quit IRC (*.net *.split) [03:07] *** ant has quit IRC (*.net *.split) [03:07] *** meingtsla has quit IRC (*.net *.split) [03:10] *** up_the_irons has joined #arpnetworks [03:10] *** toddf has joined #arpnetworks [03:10] *** twobithacker has joined #arpnetworks [03:10] *** RandalSchwartz has joined #arpnetworks [03:10] *** brycec has joined #arpnetworks [03:10] *** tabthorpe has joined #arpnetworks [03:10] *** ant has joined #arpnetworks [03:10] *** meingtsla has joined #arpnetworks [03:10] *** sinisalo.freenode.net sets mode: +oo up_the_irons toddf [03:12] *** davantalus has quit IRC (*.net *.split) [03:12] *** awyeah has quit IRC (*.net *.split) [03:13] *** davantalus has joined #arpnetworks [03:13] *** awyeah has joined #arpnetworks [03:14] *** NiTeMaRe has joined #arpnetworks [04:39] *** twobithacker has quit IRC (Read error: Connection reset by peer) [04:40] *** twobithacker has joined #arpnetworks [07:29] *** fink has joined #arpnetworks [07:31] anybody have experience with vultr? [07:31] https://www.vultr.com/pricing/ [11:11] fink: According to the IRC logs, yes some people do http://irclogger.arpnetworks.com/irclogger_log_search/arpnetworks?search=vultr&action=search&error=0 [11:12] brycec: cool, thanks [11:13] np - I knew I'd heard the name recently [11:13] i'll try it out [11:14] you get a lot of ram for your buck [11:40] i tried it [11:40] cos they have sydney location [11:40] i'm using up most of the bandwidth on it :( [11:40] there's a few annoying things about it [11:40] like when doing ubuntu it installs this cloud crap [11:41] so i screwed up my bootup, when i was trying to shift away from dhcp and that auto dns manage thing [11:41] but in the end i screed up because i set netmask wrong. [11:42] they don't do fancy /30s like arp :) [11:42] so you get all these arp requests [11:43] hmm i'm seeing bootp/dhcp requests even [14:24] *** Guest9050 has quit IRC (Quit: ZNC - http://znc.in) [15:06] *** erratic has joined #arpnetworks [15:06] *** erratic is now known as Guest62184 [16:50] *** dj_goku has joined #arpnetworks [18:05] mercutio: do they put a whole bunch of hosts in a /24 ? [18:05] ie: different customers [18:09] some of them are /23s [18:09] but yes. [18:09] it's actually very common [18:53] *** tehfink has joined #arpnetworks [18:57] *** fink has quit IRC (Ping timeout: 265 seconds) [18:57] *** tehfink is now known as fink [19:24] If I had a supermicro server with 2 nics for connectivity, as well as one IPMI nic for power cycling / remote console connectivity, how would I setup access if I only have one ethernet cable from my provider that comes to my 2U colocation space? Would I use something like a Cisco ASA Firwall to protect the IPMI interface, while providing connectivity to the 2 regular Nics ? [20:11] *** fink has quit IRC (Quit: fink) [22:22] something that can do transparent bridging [22:22] or just a switch would work, if you have some other way to restrict traffic to the management nic [22:29] For that matter, perhaps the BMC has its own firewall [22:44] mnathani: lots of them can run the ipmi through the primary ethernet [22:45] but ipmi's aren't really safe on the internet [22:45] the bmc's suck, and supermicro suck just as much or more than other.s [22:45] (As most embedded devices aren't, due to lack of update) [22:45] ip acl would be cool i suppose [22:46] i'm not aware of being able to do that easily [22:46] you can set vlan [22:46] but providers don't seem to usually provide vpn access to oob [22:46] arp does :) [22:47] normally i'd jkust recommend to run 2 1u servers rather than 1 2u, and be able to bounce via the other one and run them redundant/backup