mercutio: http://pastebin.com/TrVyaw3L
http://pastebin.com/L5g7Qswt
http://pastebin.com/jwkqCXHB
seems my home server is by far the fastest
but the first one is xen
i think this test is mostly just cpu and system overheads ***: kevr has quit IRC (Quit: ZNC - http://znc.in)
kevr has joined #arpnetworks
kevr has quit IRC (Quit: ZNC - http://znc.in)
kevr has joined #arpnetworks mercutio: hmm i only had to enable cache for java to make ipmi java kvm thingy work ***: Erick- has joined #arpnetworks
Erick- has left
dj_goku has quit IRC (Remote host closed the connection)
mordacw has joined #arpnetworks
sga0 has joined #arpnetworks mordacw: Is garry in europe, or just a workaholic? He was setting my VPS up at 2am PST. ;) mercutio: heh
he just works late. mordacw: cool. brycec: workaholic, yep
Actually mordacw, he's in Cali like you :P RandalSchwartz: at least you can honestly use PST now.
drives me nuts when people use that during the summer. brycec: Ditto
If someone write PDT in June, I'm showing up an hour off mercutio: skype seems to do something like that with my timezone
on windows hah. but it seems to have it's own time zone idea and tell people it's a different time than it is for me.
although it says it's daylight saving when it's not. RandalSchwartz: brycec - why an hour off?
or was that a typo
PDT in June is correct brycec: Oh right, right mercutio: hahahaha brycec: my point remains...
in spite of my typo mercutio: i think daylight saving should just be done with . RandalSchwartz: yeah, and when I have to schedule international skype calls, it's really crazy
because of stuff like this: http://www.timeanddate.com/news/time/
constant fiddling around with time zones. mercutio: international skype calls can be hard for me even without time zones changing
california is one of the easiest though, and it's even easier with daylight saving shifting difference to 21 hours.
europe on the other hand..
hmm it seems a few areas are moving to daylight saving all the time RandalSchwartz: 21 hours for dst difference?
yeah, year-round dst doesn't make sense to me
why wouldn't you want the sun mostly overhead at noon. mercutio: well it was 19 hours a few months back RandalSchwartz: ahh.. so it's 19, 20, or 21 then mercutio: yeah RandalSchwartz: unless your dst shifts are alighned mercutio: but I think of it was 3 to 5 hours difference
nah not aligned.
i think of it 3 to 5 ahead, and a day behind.
so it'd be tuesday 11:56 am atm
but the weird thing i find, is with europe etc, it's kind of disconcerting talking to someone when it's your morning and their evening.
i don't see why you'd want the sun overhead at noon myself.
but it's often hotter here at 1 to 3 pm brycec: After the return to PST, the sun is at "high noon" around 11am :/ RandalSchwartz: well - tradition mercutio: and when daylight saving starts people tend to get up in the dark BryceBot: get on up
and DANCE -: BryceBot dances :D-<
BryceBot dances :D|-<
BryceBot dances :D/-< mercutio: for a gun fight RandalSchwartz ?
https://letsencrypt.org/ -: RandalSchwartz sets up letsdecrypt.org mercutio: heh
well it's interesting at the least.
% whois letsdecrypt.org
NOT FOUND BryceBot: That's what she said!! mercutio: you should register it :)
my main arguement against encryption everywhere is caching, but if there was secure connection to proxy that then does secure connection to web sites i'd be happy with that
and if it shows rather than being captured mitm ***: mordacw has quit IRC (Quit: 403 attention redirect) up_the_irons: mordac: i'm a workaholic
and i tend to find that if i work late, more gets done (less distractions) ***: kevr has quit IRC (Quit: ZNC - http://znc.in)
kevr has joined #arpnetworks RandalSchwartz: less distraction... or perhaps fewer distraction*s* ***: dj_goku has joined #arpnetworks up_the_irons: are you grammar raping me RandalSchwartz ? mordac: up_the_irons: any chance you can add openbsd 5.6 to the ISO list? RandalSchwartz: I think you mean "rapping". acf_: does OpenVPN just have really sucky performance in general?
I'm having trouble getting > 200 mbit/sec with encryption disabled
on a 3.0GHz Pentium IV mercutio: it could do
you may get better performance with l2tp
but i'd avote for upgrading from a pentium 4 if at all possible acf_: yea, probably going to have to mercutio: l2tp is kernel-based. acf_: I looked at that a bit
I know I get super good performance with kernel-based GRE mercutio: it's pretty easy to setup on linux, freebsd, openbsd. up_the_irons: mordac: yeah i'll do it soon. can you give me a URL to curL? mercutio: up_the_irons: it's on your server :) up_the_irons: mercutio: where? :) mercutio: http://mirrors.arpnetworks.com/OpenBSD/5.6/amd64/install56.iso up_the_irons: danke mercutio: you may want to include the i386 version too
i dunno which mordac is running up_the_irons: i will mercutio: it's just s/amd64/i386/ up_the_irons: yep mordac: up_the_irons: http://ftp5.usa.openbsd.org/pub/OpenBSD/5.6/amd64/install56.iso
up_the_irons: http://ftp5.usa.openbsd.org/pub/OpenBSD/5.6/i386/install56.iso
redwood city mirror :) mercutio: oh ftp5 is back?
mordac: arp has it's own mirror :)
105 MB/sec hmm up_the_irons: they're up there now, i'm syncing them to each server now
ok, should be ready mercutio: ahh i assumed you would nfs them for some reason up_the_irons: nope mercutio: better to have local anyway :) up_the_irons: i don't want a network problem to cause VMs to hang
when local storage is cheap enough anyway mercutio: oh true i'd just think it'd be installing, but maybe some people leave them mounted.
yeah up_the_irons: KVM always tries to stat the ISO file when a VM boots mercutio: yeah nfs is nasty :) BryceBot: That's what she said!! mordac: up_the_irons: thank you up_the_irons: mordac: np mordac: up_the_irons: i/o load may be high on this vps. copying files off the cd is stalling periodically. it's making progress. this is not an problem for me, personally. it'll get there.
s/vps/vps host/ BryceBot: <mordac> up_the_irons: i/o load may be high on this vps host. copying files off the cd is stalling periodically. it's making progress. this is not an problem for me, personally. it'll get there. mordac: it's copying at like 11kB/s, if that's helpful. up_the_irons: i'm not sure which host you are referring to mordac: Sorry if unclear: I mounted the obsd 5.6 CD into my VPS. I'm copying files. It's copying at 100kB/s. Since it's on-disk, that seems really slow. However, it's not at all a problem for me, but since you're the admin, it might be a point of data that is relevant to you. up_the_irons: i thank you for the information. what is your VNC host? (this will tell me on which host your vps runs)
so i can check out the load right now and see what's up... mordac: kvr09 up_the_irons: cool
hmm... not too bad
18:31:35 up 188 days, 18:03, 2 users, load average: 2.05, 2.14, 1.71
that's pretty average... maybe IDE emulation is slow on KVM/QEMU.. dunno mordac: I did this same thing off the 5.5 CD earlier today and it was moving at 1-3MB/s -: mordac shrugs up_the_irons: hmm. quite a bit faster -: up_the_irons shrugs too mercutio: ide emulation is slow on kvm
virtio made a huge diff
it's not 100kb/sec slow normally though mordac: i wonder what WCHAN my host process is in
and if process is in d-state a lot
the upgrade is done, anyway. :) mercutio: oh you were doing an upgrade?
you can just use bsd.rd you know? :)
you just wget the new bsd.rd and do http install mordac: yeah. i'm a creature of habit :) mercutio: 5.6 is a bit of a pita
you can't run 5.5 binarys. RandalSchwartz: but you can watch 2 and a half men. :) mnathani: mercutio: here is the full unix bench of that liquidweb dedi: http://pastebin.com/0bbM7Ys6 mercutio: i kind of decided unixbench was crap BryceBot: That's what she said!! mercutio: well not representitive of real world performance mnathani: have an alternate I can run? mercutio: nope
i don't actually know any good ways to test performance mnathani: have you tried sysbench mercutio: even a kernel compile is a better test in some ways
but like that test shows virtualisation having way more of an impact than it does in real world
and it makes ssd performance seem no better than hard-disk performance
like my ssd raid system is getting slower file results :)
i suppose you'd want to test php and mysql and so on for a better test
benchmarking is quite a complex issue in some ways, .. and things like static file serving performance matter very little in the real world.
but dynamic can matter.
like for most web sites whether you can do 23000 or 26000 static pages/sec makes 0 dfif.
but dynamic page performance can maek a diff
and like for i/o it's not really burst perofmrance that matters, it's how slow it gets
that said, i have no idea why my ssd system benchs disks slower, it has faster cpu.
and all 3 systems i tested were mdadm raid 10 for /
i suppose a good test would be some kind of chroot package of php/mysql/nginx or something
even ssd vs hard-disk in real world mostly benefits write-heavy work loads. as usually reads can be cached in ram, as most people have datasets smaller than ram that are hot. mnathani: they have good benchmarks for desktop windows systems specifically gaming benchmarks mercutio: yeah mostly people focus in gaming / 3d benchmarks.
which don't tend to show ssds benefitting as much as they do too
my windows system is slow :(
chrome seems to be getting slower over time.
but chrome benchmarks high. it's probably my profile or something. mnathani: http://www.passmark.com/download/pt_download.htm try that one on Windows mercutio: actually chrome is slow on my linux box too, but that's a diff issue i think
i found a display fps thing, and it onyl ever seems to go up to 30 fps
i didn't see anything to disable vsync.
yeh i've tried passmark before too, it's kind of useless.
it's way too synthetic.
i mean it gives a basic idea.
linux chrome seems slowest at scrolling pages which i suppose isn't benchmarked. ***: dj_goku has quit IRC (Remote host closed the connection)
joepie91 has quit IRC (Ping timeout: 256 seconds)
joepie91 has joined #arpnetworks
kevr has quit IRC (Quit: ZNC - http://znc.in)
kevr has joined #arpnetworks mnathani: mercutio: do you do any hardware RAID on your systems? mercutio: nah mdadm+zfs
actually i use hardware raid on openbsd.
software raid is generally better than hardware raid these days ***: kevr has quit IRC (Quit: ZNC - http://znc.in)
kevr has joined #arpnetworks mercutio: also you can do fancy things like raid 10 on 3 hard-disks
and split raid10/raid5 brycec: And soft-raid is a lot more portable between machines
(such as if the raid controller itself has failed... )
(in which case you need to find another of the same controller or the existing raid volume will probably not be recognized) mercutio: i actually hate mdadm :)
the performance is good, but it can be a bit weird about failing drives.
but zfs is amazing
pity it can't be included in linux by default :( hazardous: hi nerds!
up_the_irons: do you care a lot about ip justification / is VPN or SSL/TLS proxying/termination acceptable justification for 5 IPs, one of my friends needs to basically proxy traffic from another host in LA to reach china telecom
via arp it's 180ms, direct it's about 300ms over cogent or he.net or something ***: dj_goku has joined #arpnetworks
dj_goku has quit IRC (Changing host)
dj_goku has joined #arpnetworks mnathani: eww cogent
can he.net reach cogent over ipv6 yet? (and vice versa) brycec: I can, yes
Looks like a direct handoff too
oh wait, nevermind
I didn't let mtr finish its thing...
From ge2-6.core1.sea1.he.net to "???" mnathani: I wonder if cogent and he.net specify in their terms of service: IPv6 transit to all IPv6 hosts, except he.net/cogent respectively mercutio: how did the route go before? ***: dj_goku has quit IRC (Ping timeout: 250 seconds) hazardous: ^ lol
i really really hate dealing with asian traffic
i have a box somewhere else in LA
it's 300-600ms, ~20% packetloss to any china telecom ip mercutio: isn't packet loss the norm for china traffic? :) hazardous: and there are actually times like peak/offpeak where its completely unusable
not really on arp, 180-200ms <1% loss
it's actually tolerable ***: dj_goku has joined #arpnetworks
dj_goku has quit IRC (Changing host)
dj_goku has joined #arpnetworks mercutio: i think sometimes there is icmp loss when normal traffic gets through
i knew someone who was on fibre(?) in china and they got 10 megabitish
but if talking to them on skype and they browsed us web sites skype woudl break up hazardous: i have to detect chinese ips and use different templates for them
if you use google cdn for jquery or google web fonts or etc
your whole page is generally blocked mercutio: i think there's some weird magic or something hazardous: or the js is injected with malicious scripts mercutio: so it's doing deep page inspection? hazardous: it's always done
gfw is basically line-rate DPI by american hardware
south korea's censorship/gfw system is far shittier mercutio: it's american hardware? hazardous: can a huawei even do 10mbit dpi without catching fire? ***: dne_ is now known as dne mercutio: i heard huawei routers were good.
they're getting popular here. hazardous: i've only seen them in conjunction with ridiculous vulnerability reports or CVEs mercutio: juniper and huwaei are both pretty popular here.
at medium to high end
i have no idea about the vulnerabilities on them.
i think they're all bad? up_the_irons: hazardous: for 5 IPs, that is fine
if he wanted a /24, i'd say gtfo ***: pyvpx has quit IRC (Remote host closed the connection)