[00:03] Wow, Linux really does have a faster network stack than OpenBSD. Wheezy on this Metal box gets at least 800mbps consistently. [00:03] 100MB test file downloaded in 1 second :p [00:07] * brycec hates network/Internet mirrors that can't keep up with his own connection [00:16] brycec: it does bug me a little :) [00:16] i find my home mirror pretty fast :) [00:16] well proxy cache, with ssd. [00:17] i think a lot of the time it's about small file performance. [00:17] and so having a low ping, and cached on ssd or ram helps. [00:18] saturating gigabit is quite hard for a hard-disk with lots of small files. [00:18] I notice it most with Sourceforge mirrors, many of which barely crack 10mbps [00:18] and raid can't really fix it. using 15k sas disks or something would help more than raid probably. [00:18] There's no excuse for that in this day and age. [00:18] oh sourceforge has it's own issues i think. [00:18] (*its) [00:18] i get 200 to 300k/sec or something on sourceforge frequuently [00:19] actually the worst i find are motherboard wsites. [00:19] if you ever download anything from a motherboard site i tend to find it can be like 20 to 30k/sec or less! [00:19] Heh I don't run into too many of those typically, but yeah they are usually slow. I don't mind much, they aren't running a professional download service :p [00:19] and it's not just because they're in taiwan. [00:19] well sourceforge don't charge for downloads. [00:20] and don't have much of a revenue stream? [00:20] i find mirrors.kernel.org seems to have consistent speeds. [00:20] well when it goes to the right mirror hah [00:20] Indeed, quite reasonable. It's my goto mirror [00:20] but it doesn't tend to get slow at diff times. [00:20] whereas archive.ubuntu.com/security.ubuntu.com can crawl [00:21] whenever there is a new ubuntu release it shows on my smokeping :) [00:21] i actually use arp's mirror a lot :) [00:21] plus nz mirror. [00:21] the nz mirrors are faster than arp it seems (when similar ping) [00:22] probably just better raid config, or more users so more likely to be in cache [00:22] haha [00:22] I'm using ARP's mirrors quite a bit right now too... on-network. [00:23] I'd get much more use out of it if ARP would just mirror Debian too. [00:23] (and stop wasting space mirroring architectures they themselves don't support) [00:24] yeah i think he's short on space. [00:24] yeah that'd be sensible. [00:25] Ditto - he's commented on it before. [00:25] Why do we mirror powerpc ISOs for FreeBSD? :p [00:25] maybe you should volunteer to clean it up. [00:25] we/they [00:25] I didn't know up_the_irons was looking for volunteers [00:25] i have no idea if he is or not :) [00:25] the full Debian archive is ~1TB :/ [00:26] I looked into downloading it once.. [00:26] I've run company mirrors in the past, I have some experience doing so :) [00:26] but i think he's kind of busy with other things. [00:26] * brycec remembers when Debian was 500GB [00:26] you probably really want 4x3tb disks or something for a large mirror. [00:26] He does indeed seem to be quite busy [00:26] if you used zfs and had a ssd cache it could be amazing :) [00:27] i was going to setup an openbsd mirror before. [00:27] I did that last month :) [00:27] (internal only) [00:27] how big is it now? [00:27] * brycec goes to look [00:28] 472GB? [00:28] hangon this site says debian is 578GB [00:28] (my vpn seems to be down, ugh) [00:28] oh there's debian CD too which is 327GB [00:28] http://mirrors.gigenet.com/ [00:29] this mirror seems to list sizes [00:29] even postfix is 1.3gb haha [00:29] https://www.debian.org/mirror/size [00:30] oh mirorring just amd64 is only 100gb [00:30] amd64+i386+source would be better [00:30] yea, that should be fine [00:32] I'd be very happy about a Debian mirror at ARP [00:32] I have lots of boxen connected to ARP for IPv6 via a tunnel [00:32] arch is only 41gb [00:32] +7gb for iso, +28GB for source. [00:33] mercutio: My OpenBSD mirror is 738GB, only mirroring amd64 and i386, releases 4.0 through 5.6 as well as snapshots, Open{SSH,NTPD,BGPD}, LibreSSL, as well as M:tier's openup [00:34] brycec: arp doesn't mirror that many versions [00:34] Okay. And? [00:34] that's quite large. [00:34] twss [00:34] Okay! twss! 'that's quite large.' [00:34] 732GB :( [00:34] http://sprunge.us/fIVN [00:35] * brycec meant to include snapshots in that too - 55.7GB [00:35] hmm [00:35] i wonder how much disk arp's mirroring is using. [00:36] http://www.amazon.com/dp/B00LO3KR96 [00:36] Amazon: "WD Red 6TB NAS Hard Drive: 1 to 8-bay RAID Hard Drive: 3.5-inch SATA 6 Gb/s, IntelliPower, 64MB Cache WD60EFRX" [00:36] And yes that's quite a few releases. It spans our development of OpenBSD-based systems, so we kinda need to have access to those old versions. (which is hard to come by, these days) [00:36] eww ed :) [00:36] wd [00:36] yeah i've had that issue before brycec [00:37] actually i'm surprised arp doesnt' go back further [00:37] it's only got 5.5 and 5.6 [00:37] ARP is just rsync'ing from upstream, which doesn't carry older than that [00:37] i nearly bought those cheap external seagate drives. [00:37] oh ok [00:37] i think it was $140 for 5tb or something? [00:38] http://www.amazon.com/Seagate-Expansion-Desktop-External-STBV5000100/dp/B00JT0EGPW/ref=lp_1254762011_1_2?s=pc&ie=UTF8&qid=1428564996&sr=1-2 [00:38] Amazon: "Seagate Expansion USB 3.0 5TB Desktop External Hard Drive (STBV5000100)" [00:38] woah [00:38] so you strip them out of their cases... [00:38] yea [00:38] yeah it's pretty good value [00:39] but seagate have some weird screw stuff with some of their larger drives. [00:39] so they won't necessarily work with trays properly. [00:39] i've got a dying disk on my personal server :( [00:39] is that a 3.5" disk? [00:39] yeah 3.5" [00:39] i was hoping to upgrade to 6tb when i got disk failure [00:40] but i'll probably just stick another 3tb in [00:40] actually zfs registered no errors. [00:40] so maybe it fixed itself. [00:40] [689969.969725] blk_update_request: I/O error, dev sdd, sector 4314598592 [00:40] you'd /think/ zfs would have an issue there [00:41] that's always scary.. [00:41] using RAID? [00:41] yeah raid-z [00:41] good :P [00:41] it has 2.6tb of disk free. [00:41] but it's in nz. [00:41] is the ARP mirror a dedicated box? [00:42] don't think so [00:42] darn [00:44] i wish amazon wouldn't combine reviews [00:46] so amazon charge $73 US for 1tb, $83 US for 3tb. [00:47] oh it's cheaper for normal 1tb, that was retail, some weird link thing [00:51] hey Brycec you could run a mirror :)) [00:51] i wonder if arp would let you provide your own hard-disks [00:52] i can't figure out if this hard-disk is the weird screw kind of not though [01:06] bit's smr [01:06] so it's randomly slow [01:28] *** mnathani_ has quit IRC (Ping timeout: 244 seconds) [06:56] you'd have to have a dedicated to have your own hard disk [06:56] and then what if it fails? [09:03] the odd# TB disks tend to have longevity issues [09:03] 1.5, 3, etc [09:04] because they do weird stuff to achieve that density [09:05] so i guess not odd numbers as in 1,3,5 etc, but odd numbers as in half numbers or combinations of half numbers [09:11] https://www.backblaze.com/blog/hard-drive-reliability-update-september-2014/ oddly enough, it seems like backblaze did a study of sorts w/failure rates [11:41] I don't find it odd that a company that consumes hard drives by the truckload did a study :P [12:28] nothing odd about it [12:28] that was the joke [12:29] at any rate, not the first time i've seen data saying the weird in-between sizes are unreliable. i just couldn't find the original one i read a while back (had a lot more specific data) [12:40] Oh I didn't catch the joke/sarcasm :p [12:41] there was an amusing post on their blog from the middle of the disk shortage that happened following the flooding in thailand [12:42] where their cost for standalone consumer disks doubled (or more) so they resorted to buying external drives and "shucking" them from their enclosures [12:42] https://www.backblaze.com/blog/backblaze_drive_farming-2 [12:58] Yeah I remember [12:58] My company was doing the very same thing [12:59] In smaller quantities at least. But we were doing it too. And having friends+family buy-out stores and ship the drives to us [13:14] *** poulsen has joined #arpnetworks [13:25] m0unds: sometimes it's hard to know what they're doin ghtough [13:25] like 2tb disks can be 2 or 3 paltters with seagate. [13:25] Easy to tell by weight and construction, though that usually comes after you've opened store packaging ;P [13:31] i thought 1tb were good myself [13:31] That's what she said!! [13:33] m0unds: you know there's iossues with that study btw? [13:34] shit i so i can't type before coffee. [13:34] basically they don't have proper vibration protection at blackblaze, and they';re using consumer hard-disks which don't uusually have good vibration tolerance as serve rhard-drives. [13:35] I'm pretty sure BB do something about vibrartion stuff [13:35] *vibration [13:35] oh? [13:35] Seen it mentioned in their detailed builds [13:36] https://www.backblaze.com/blog/backblaze-storage-pod-4/ [13:36] "" [13:36] "Includes case, anti-vibration assemblies, cross supports, etc." [13:36] the only recent disk failutres i've had have been with wdc black 1tb and seagate 3tb [13:36] but that doesn't say much on it's own :) [13:37] (I haven't actually seen their writeup on "4.0" before, but it was the first google result) [13:43] our failure rates w/ES2 and 3 seagates was less than 3% [13:46] hard-disk tech doesn't seem to have improved recently [13:46] dynamic spindle speed stuff is kind of nice for energy use reduction [13:46] they're doing some "density improvements" but it seems it coudl be at the expense of reliability/performance. [13:47] yea [13:47] i'd be happy not buying another platter disk again [13:48] i'd like cheap bulk ssd that is slower [13:48] and to have fast ssd cache in front of it :) [13:48] i'd really like to justify these 2gb/sec ssd's. [13:48] but motherboards with m.2 are really expensive. [13:49] you can get pci-e adapters, but they require a pci-e slot... [14:02] i wish intel's chipsets would have more pcie lanes to share [14:02] and shift to pci-e 3 [14:03] amd provided 48x of bandwidth starting w/their 990fx chipsets [14:03] of pci-e 3? [14:03] whatever the spec was in 2010 when that chipset came out [14:03] you could run a gpu at 16 and a nic or sata controller or whatever, and neither would downgrade [14:04] my onboard ethernet is only 2.5GT/s [14:04] 1 lane [14:04] or two gpus and a nic and still have them at native speeds. only downgrade occurred w/4 gpus + anything else [14:04] where my 10 gigabit is 8 lanes at 5GT/s [14:04] so has 16x as much bw [14:04] i'm wondering how it'd go in a 2x slot [14:05] on my board w/z87 chipset, 1 gpu + 1 nic + 1 other device = gpu at 4x [14:05] most dual/quad gigabit cards are 4x. [14:05] does gpu at 4x hurt? [14:05] yep [14:05] doesn't really at 8, does for sure at 4 [14:05] for 2d? [14:05] nah, 3 [14:05] 3d [14:05] i'm at 8x/8x on my windows box [14:05] my pc's primary job is entertaining me with explosions and such [14:06] and 16x/4x on my linux box [14:06] but the 4x is chipset :( [14:06] i like 8x/8x [14:06] yea, 8/8 is fine for the most part, esp w/pcie3 [14:06] so video card moving to 4x pci-2 may hurt? [14:06] yea [14:06] even for 2d? [14:06] doubt for 2d [14:06] for 3d yes [14:06] hmm [14:06] what about chrome [14:07] couldn't tell you [14:07] hw rendering is chrome breaks pretty regularly on every platform [14:07] i dunno i'm replacing motherboard anyway. [14:07] s/is/in [14:07] hw rendering in chrome breaks pretty regularly on every platform [14:07] i used to find linux had amazing chrome performance compared to windows [14:07] and i managed to improve my web performance with chromium on windows. [14:08] so recently i tried my old chromium build, a whole lot of sites complain about old version, but it seems way faster. [14:08] i was mostly screwing around with allowing more than 16 connections through http proxy iirc [14:09] i had all these annoying if-modified-since requests going through to proxy [14:09] and you can't just throw 30 if modified sinces out at once :( [14:09] and get a list of what's modified [14:10] i hope http/2 fixes that kind of case, it was pretty common [14:11] it's kind of well known that web pages will load faster if you rename images, (use unique names) and set long expiriation times to avoid if-modified-since [14:11] but hardly any real sites do that. [14:19] *** mnathani_ has joined #arpnetworks [16:30] so i'm doing business cards and a website for my music stuff [16:30] i'm thinking i should probably just go with a wordpress install for the music site [16:30] the people that are going to help me with content aren't exactly tech experts [16:31] but i want to put video files and an audio player on it [16:31] so people can pick and choose what to listen to..that sort of thing [16:31] no idea how to do that portion [16:32] Interesting. [16:32] phlux: Take a look at Concrete (C5) too [16:32] theres porbably a plugin for multimedia content [16:32] It aims to be more user-friendly, and more secure [16:32] uploading and viewing / listening [16:32] s/porbably/probably [16:32] theres probably a plugin for multimedia content [16:34] I wonder if there's a good multimedia addon for Concerte [16:34] Concrete, even [16:36] Something like this: https://www.wonderplugin.com/wordpress-audio-player/ [16:57] *** ix33 has left [17:01] *** ix33 has joined #arpnetworks [17:05] phlux: are you doing a facebook page? [17:05] yes [17:05] it seems facebook is one of the best ways for promotion these days [17:05] as you don't have to go so far out of youur way to check the page [17:06] yeah [18:19] *** poulsen has quit IRC (Remote host closed the connection) [22:05] soundcloud has an embeddable player and their free tier allows around 1hr of content