#arpnetworks 2015-03-20,Fri

m0unds

yea, around there

mercutio

i updated my smokeping with my newer targets list at least.

m0unds

happens between 0830-0900 MDT

mercutio

i'm never around then :)

m0unds

yeah

mercutio

do you have any mtr's or anything?

m0unds

i can't MTR the arp box because of how severe the packet loss is

mercutio

does it get unusable?

m0unds

yep
i can't stay connected via ssh at all

mercutio

can you setup a cron on the arp box to mtr out?
does packet loss start before it hits arps network?

m0unds

no
aside from the naturally occurring deprioritization stuff

mercutio

and it's lossy to arp's web site or other arp hosts at the same time as your vm?

m0unds

haven't tested it

mercutio

yeah i hate lboody deprioritisation :)

m0unds

i guess i will next time this happens

ihttp://pastebin.com/F7SJ0GjA

mercutio

what kvr are you on?

[13:56]

m0unds

[13:56]

mercutio

oh
yeah :)
well i'm monitoring that host
and i've only seen loss a couple of hours ago

[13:56]

m0unds

my latency to the vm is jittery as all hell
+/- 12ms

[13:57]

mercutio

even now?

[13:57]

m0unds

yep

[13:57]

mercutio

does it get unusable?i'm seeing 0.1 msec jitter or something
oops
i must have pressed up arrow a bit
yeh no jitter here
well a mtr now could be useful.
i usually use mtr -i=0.1 when i'm testing things :/
which shows up deprioritisation even more
err mtr -i0.1
but shows jitter easier
r1 drops packets with that frequency

[13:58]

m0unds

[14:02]

mercutio

wow
oh you traced to kvr03 :)

[14:03]

m0unds

yeah
tracing to te vps now
it's largely the same

[14:04]

mercutio

that was with -i1 ?
it could be a comcast issue

[14:04]

m0unds

negative
it's not

[14:05]

mercutio

[14:05]

m0unds

no packet loss to any other destination i've tested

[14:05]

mercutio

ther's jitter to 96.120.1.73

m0unds

buddy in FL was using the voice server and started noticing packet loss and let me know

mercutio

but i mean between arp/comcast.

m0unds

yes, because this is a residential cable connection

BryceBot

That's what she said!!

m0unds

BryceBot: no

BryceBot

Oh, okay... I'm sorry. 'yes, because this is a residential cable connection'

mercutio

but there' snot jitter to the hop prior of 689.86.182.122
but if 96.120.1.73 is a bras, that could be deprioritisation

m0unds

it's the headend
so yes

[14:07]

mercutio

yeah
mind sharing your ip? :)

[14:07]

m0unds

73.26.90.138

[14:07]

mercutio

it's worse to your ip
i'm still seeing no loss though
and it's just as bad from another provider

[14:08]

m0unds

my latency to a box w/another provider is 30.1ms +/- .2ms

mercutio

oh but it's fine from new zealand
well 3 msec jitter
or just under

BryceBot

That's what she said!!

m0unds

BryceBot: no

BryceBot

Oh, okay... I'm sorry. 'or just under'

mercutio

can you try tracing to 202.49.71.24 ?
that's my nz test server

m0unds

i have less jitter to that destination than to arp

mercutio

it's gtt->comcast for forward route

m0unds

+/- 1.8ms

mercutio

but i suspect it's forward path from youu that's the issue
yeah i know right
i imagine it's not going comcast->level3

m0unds

but that doesn't explain why a bunch of other peoples' connections took a dump around 30 seconds after i started seeing packet loss

[14:11]

mercutio

they may all be on comcast?

[14:11]

m0unds

congestion could explain the jitter, but i couldn't care less about jitter
no, their sessions to freenode dumped

[14:11]

mercutio

but yeah level3 may be having issues
oh hmm

m0unds

on arp V6 IPs (i saw at least one with no rdns that was arp's prefix)

mercutio

it's not even peak time though
true
the mind boggles :)

m0unds

that's why i'm wondering if someone's machine is doing something at that time
because it happens right around that time each time

BryceBot

That's what she said!!

m0unds

BryceBot: no

BryceBot

Oh, okay... I'm sorry. 'because it happens right around that time each time'

mercutio

yeah

m0unds

like maybe saturating the link to upload or download something

mercutio

i don't think kvr03 is having issues
and i don't think it's likely you're all on kvr03
and i don't think r1 is having issues
but s7 coudl be having issues
it doesn't really explain ipv6 screwing up thouguh

m0unds

nope

[14:14]

mercutio

actually it does
ntt is on s7
if it's a general issue

[14:15]

m0unds

so, having looked at the users w/out cloaks who dropped when i mentioned packet loss, they were all connected via some arp svc

[14:15]

mercutio

yeah
incoming skips s7
for ipv4 from r1
and my r1 based routes have been fine
but outgoing doesn't skip it

[14:15]

m0unds

hey, milki jpalmer RandalSchwartz awyeah - are you guys connecting to IRC via an ARP VPS or dedi? if you're on a VPS, mind sharing which host?

[14:17]

mercutio

but you wouldn't think someone would ddos at the asme time every day

[14:17]

m0unds

right, that's why i'm thinking it's something saturating a link
or similar

[14:17]

mercutio

and downloading at high speed over transit enough to saturate links isn't normal

m0unds

offsite backup, something like that

mercutio

yeah it could be that they're pushing backups onto arp

m0unds

if they all happen to be on kvr03, that'd be great
haha

mercutio

on a dedicated server
i really don't think kvr03 is at fault
but i'm happy to be wrong
oh hmm
kvr03 still goes throuugh s1
i was wondering if hitting a host directly would bypass s1

m0unds

outbound route from the vm on kvr03 is via s7

[14:20]

mercutio

and that may happen on peering but not transit

***

hoggworm has joined #arpnetworks

mercutio

inbound route is via s7 too i think you'll find
it just responds with the link address to level3 probably
but it'll be the same host
it's only r1 that's assymetric
because s1 doesn't have full route table

m0unds

oh, were you meaning tracing the host kvr03.arpnetworks.com indicates s1?

mercutio

well i was meaning kvr03 may have more direct path than your vm
but nah it's the same path ish

[14:22]

m0unds

yeah, looks the same
it's pretty symmetrical. much better than before level3 was turned on

[14:22]

mercutio

yeah level3 isn't bad

BryceBot

That's what she said!!

m0unds

used to be inbound from nlayer/mzima/whatever and outbound via ntt

mercutio

nlayer would have been fine probably

m0unds

and it was unusable during "peak" time for me
yeah, it was NTT that sucked
way way way oversubbed at comcast

mercutio

nlayer's been having weird shit happen recently
nlayer/mzima/tinet/gtt
i can't tell which is which haha

m0unds

yeah, hahah

mercutio

as4436 is having issues
which is that

m0unds

mzimtinettttttt

mercutio

it says nlayer :/

m0unds

yeah, some stuff still does

mercutio

http://www.peeringdb.com/view.php?asn=4436
yeah so that's the legacy nlayer

m0unds

the net formerly known as nlayer

[14:25]

mercutio

what happened to mzima haha
i wonder how many routes comcast has
245
including ipv6
there's some non aggregated stuff in there too
i imagine they have more asn's, that's just AS7922
it kind of sucks that comcast are huge and don't peer :/

[14:25]

m0unds

they started splitting up their network a bit more
based on market and stuff, tons of ASNs

[14:34]

mercutio

ahh
probably a good thing

[14:34]

m0unds

yeah

mercutio

mkb isn't on comcast is he

BryceBot

That's what she said!!

mercutio

as you said it doesn't seem like it's just a comcast issue
but prioritising making sure users connections well seems sensible
freenode on the other hand..
dropping on freenode could be anything
that said, i've been stable for a whil again

does this packet loss happen at toher times than 7:30 to 8:00 MDT?

m0unds

it has before, yes
it happened a ton last week (i think it was last week)

mercutio

but atm i mean
yeh i remember that

m0unds

well, it hasn't happened since last week

mercutio

ahh

[14:45]

m0unds

today's the first day. the time when it occurred in the past started at 0830-0900 MDT and then would occur again throughout the day

[14:45]

mercutio

that gave me loss though
this one didn't give me loss
also kvr03's reoccuring issues in the past got dealt with
the specific to kvr03 issues
which was fixed about 10 days ago
but there was the issue efecting everyone around then too
and the effecting everyone was never determined the cause of i think

[14:45]

m0unds

borrowing smokeping data from acf_ - http://kremvax.acfsys.net/smokeping.cgi?epoch_start=1426863600;hierarchy=;epoch_end=1426888048;target=Remote.arpnet-lsanca;displaymode=n;start=2015-03-20%2008%3A00;end=now;Generate%21=Generate%21
it's on there

[14:49]

mercutio

ahh i see

[14:50]

m0unds

0750 PDT 0850 MDT

[14:50]

mercutio

it's level3 outbound
to ec2
i wonder if it's level3 having issues

[14:50]

m0unds

i think this is local though
microsecond ping time
on that graph, avg is .688ms

[14:51]

mercutio

oh
nah his smokeping is weird
the top is to arpnetworks.com
but the last 30 hours is to ec2.acfsys.net

[14:51]

m0unds

yeah, from the local box
second is from ec2

[14:52]

mercutio

it's the second that's having issues thoguh

[14:52]

m0unds

the graph i linked, top part, showed 30+% packet loss to arpnetworks.com from within the network
for a period of 10 mins

[14:52]

mercutio

oh the link didn't work

m0unds

mercutio

i dununo why

m0unds

weird

mercutio

it wa so long, and it didn't make sense why

m0unds

ok, well set the time today (03/20/2015) and 0700 as the starting time
to now
http://i.imgur.com/Z5WZ0y4.png as an alternative

mercutio

oh that's better
mine was set to 08:00

m0unds

yeah

mercutio

and it was ll grreen :/

m0unds

that weird little blip is there, even on arp>arp stuff

mercutio

you're so right
http://la.meh.net.nz:24/cgi-bin/smokeping.cgi?target=meh.arpmeh
that's to arp vm from vm on arp dedicated
hmm
maybe it is kvr03 issue?
i can't rmeember what host acf is on

m0unds

http://i.imgur.com/jFD4Yic.png
wishing i still had my server at home so i could have additional data
but i shut down everything here but an rpi
haha
a rpi*

[14:57]

mercutio

acf ios on kvr27
looking at logs
he may have more than one host thouugh

[14:59]

m0unds

i saw something about kvr10 - i was on 27 before (had a project that ended)

[15:00]

mercutio

oh
i saw discussion about 10 too
but he said he had high latencies to 27
but that may have been you :)

[15:01]

m0unds

yeah, we were trying to figure out the ntt weirdness
around then

[15:01]

mercutio

i should just monitor them all :/

[15:01]

..... (idle for 22mn)

brycec

Thanks for the reminder m0unds, I keep meaning to setup smokeping on my rpi
(but it still won't be today)

[15:23]

BryceBot

That's what she said!!

[15:23]

mercutio

i need to set it up on my san jose vultr host
which seeems to be ntt

m0unds

if i weren't running an awful java app on mine i would

mercutio

so useful for testing transit from arp i suppose

m0unds

i need to get another one

mercutio

the network is pretty bad on vultr

m0unds

is it?

mercutio

well it depnds
http://sj.meh.net.nz:24/10m
vs http://la.meh.net.nz:24/10m

m0unds

they're both the same for me, 5MB/sec

mercutio

maybe youu're closer

m0unds

i likely am

mercutio

vultr has random issues more often too
oh wow
from uk sj is way faster than la
ntt vs level3

m0unds

i've only used a couple hosts long term over the last several years
arp and linode for the last 2 or so

[15:31]

mercutio

i only use vultr cos they're cheap
just for smokeping and bw tests haha
and somewhre convenient to ssh to
ok this is strange
vultr is going ntt->level3
and getting better performance than arp going level3->level3

[15:32]

i wonder if something is up with level3
arp to vultr was fast.
but sj vultr shouldn't be twice the speed of arp
esp when they're both l3 to uk
(that was doing a uk test)
and vultr has the extra ntt to level3 hop

[15:40]

http://pastebin.com/HNz6uQtK
anyone want to try my curl download test?
it downloads 7 10mb files

[15:49]

***

YogeeBear has joined #arpnetworks

[15:50]

mercutio

i probably should just stick it on site
http://weallsee.net/testspeed

[15:51]

brycec

mercutio: http://sprunge.us/GdAh but 10MB is a pretty small test file for a good connection with 100mbps Internet.

[16:02]

mercutio

brycec: it's good to be small

[16:02]

brycec

(that's from my home cable connection)

[16:02]

mercutio

large fikles you usually have in background
smaller ones you have in foreground
and it gives you a good idea if connection is going wekll

brycec

But for a speed test, a 10MB file on a connection that can carry 10MB/s is a poor sample size.

mercutio

i usually time the whole script
but adding up the numbers it looks pretty good

brycec

mercutio

but la is slower than sj
why is la so slow

brycec

(I'm also closer to SJ)

mercutio

oh
but look how fast you get to new zealand
the top one is new zealand

brycec

ah
faster than .au, impressive and surprising

mercutio

la.meh.net.nz is a nz ip advertised on an arp vm
i hope it's not going via nz
nz is actually closer to the US than asutralia
and syd is vultr :/

brycec

yeah but aren't most trans-pacific cables going to Sydney directly?

[16:05]

mercutio

i thought it was more imrpessive that it's faste than los angeles
nope
well they are
but it's a longer cable
i probably should use an arp ip
but as a quick test, la.meh.net.nz ping looks fien?

[16:05]

brycec

(currently running mtr's to see)
(Looks like I get to NZ via Cogent's SJ -> Plain.net.nz direct)
(And Sydney via NTT LA / Equinix Sydney)
http://sprunge.us/GNCS if you're curious
la pings were slightly slower than san jose, and both went over NTT

[16:06]

mercutio

well it matters more what my route to you is
it's 3.18 kernel on la vs 3.19 on the others
it's gtt->comcast from nz
and it's going direct via san jose
and la is sending out via level3 to comcast.
with way worse jitter
i mean icmp deprioritisation and all that, but it's 0.3 msec jitter vs 2.1 msec jitter.
the jitter on seattle level3 is showing 0.3 msec jitter though
the jitter is on te-0-1-1-ten01.spokane.wa.seattle.comcast.net
and te-0-1-0
they both hit that hop
which alternates.

[16:13]

brycec

(I should point out that Spokane, roughly wear I live, is a city 5 hours East of Seattle)

mercutio

actually they both hit be-37-sur03.spokane.wa.seattle.comcast.net
and the jitter isn't there

BryceBot

That's what she said!!

brycec

(So, presumably Seattle serves as our upstream, but it is aways away)

mercutio

so probably it's forward path
ahh your mtrs only have one ping

brycec

Yeah, for expediency

[16:20]

mercutio

it's hard to know thouugh
yeah
can you run that test again just to make sure it's not random that la is going slow?

[16:20]

brycec

sure

[16:23]

mercutio

i mean 30 megabit isn't /that/ slow

[16:23]

brycec

do you want just la's mtr?

[16:24]

mercutio

just la's mtr
and la's download speed
actually just doing la a few times over for the 10mb test would be fine
it gets over 40mb/sec from vultr
so i doubt it's a general throughput issue
which is making me wonder if level3 is being weird

[16:24]

YogeeBear

DCC SEND STARTKEYLOGGER 0 0 0 سمَـَّوُوُحخ ̷̴̐خ ̷̴̐خ ̷̴̐خ امارتيخ ̷̴̐خ

[16:25]

***

YogeeBear has left "DCC SEND STARTKEYLOGGER 0 0 0 سمَـَّوُوُحخ ̷̴̐خ ̷̴̐خ ̷̴̐خ امارتيخ ̷̴̐خ"

[16:25]

mercutio

if it's slightly problematic generally, it increases the chances that it's being more problematic other times

[16:26]

brycec

mercutio: http://sprunge.us/LRJG
(and traffic on my own net is relatively quiet)

[16:26]

mercutio

wow
it's all ove rthe place.
i reckon there's packet loss
can you try the same thing with garlic.plain.net.nz:24/10m ?
i expect it'll be the same

[16:28]

brycec

sure, one sec

[16:33]

mnathani_

anyone use policy based routing to redirect certain traffic out a different gateway?

[16:33]

mercutio

i've done it before mnathani_

brycec

mercutio: http://sprunge.us/dMJd

mercutio

using ip rule and multiple tables

brycec

I redirect specific subnets and addresses over my work vpn, yeah

mercutio

yeah same diff with garlic
it's xen, garlic is dom0
why does your router have such high pings?
just seems a bit strange :)

[16:35]

brycec

because I don't have direct ethernet to it

mercutio

oh it's bridged?

brycec

(ethernet-coax-ethernet)

mercutio

ahh

brycec

decanet specifically
pings to other hosts on this side of the bridge are .3ms

mercutio

maybe we could try iperf udp test?

hoggworm

sup with yogeebear trying to dccc the file "startkeylogger"

brycec

hoggworm: #frenode is already aware

mercutio

it's a pita if you can't easily forward udp

brycec

hoggworm: https://nullroute.eu.org/~grawity/startkeylogger.html

hoggworm

brycec: cheers

brycec

[16:39]

mercutio

that startkeylogger thing is old apparently
and it's been in other channels earlier

[16:39]

hoggworm

oh
haven't seen it yet on ircnet/efnet/dalnet

[16:39]

mercutio

apparnelty it's an old mirc thing

brycec

I first got a random CTCP in Arabic from them :/

mnathani_

brycec: so destination address based routing?

brycec

mnathani_: just simple "route add..." stuff

mercutio

i used it for source based routing

mnathani_

I am trying to route my chromecast over a vpn

brycec

(and I use pf to limit sources)

mercutio

yeah that's easy mnathani_

mnathani_

without forcing all wireless traffic over the vpn
oh and I want to be able to stop routing over the vpn with a command

mercutio

http://blog.scottlowe.org/2013/05/29/a-quick-introduction-to-linux-policy-routing/
does that help you?

brycec

command: "Woman, unplug this cable" :p

http://pastebin.com/fDh8A1iW

mnathani_

ie: no config changes or dhcp lease changes on the chromecast

[16:42]

mercutio

of course you'll want to set a static ip or permanent dhcp ip address on the chromecast

[16:42]

mnathani_

dhcp reservation is what I have now
I am attempting to do this using vlans / cisco layer 3 switch and Router if required

[16:45]

mercutio

if you have a linux box
just set the chromecast to gateway to the linux box
and look at that link i linked for policy routing
basically youu can have two route tables
ip rule add from 192.168.30.200 lookup custom
so you can have a table called custom

[16:46]

mnathani_

would a linux vm work, or will it work out better to have a physical box with multiple interfaces?

[16:47]

mercutio

vm is fine
ip route add default via 192.168.30.1 dev eth1 table custom
then you do something like that

[16:47]

mnathani_

are these going to converto to iptable rules?

[16:48]

mercutio

hmm i did a udp test to uk nad it was fine

[16:48]

acf_

network namespaces in Linux are also nifty

[16:49]

mercutio

it's even got less jitter than vultr
so why is http going slow

[16:49]

..... (idle for 22mn)

this level3 thing just gets weirder
with multiple connections to the uk one of them goes way faster than the other
so there's some load balancing going on
so my udp test may have been fine because it was on the fine connection
so bryces huge varying degrees of speed for the 10mb file may been due to alternating load balanced connections

[17:11]

mnathani_

mercutio: how can I test the curl 10 meg files

[17:19]

mercutio

mnathani_: http://weallsee.net/testspeed
it's a shell script
with the urls
and set -x so you know which host it's going to :)
i'm seeing jitter to comcast now too
from nz
it was fine before, so comcast's load balancing is lame too :)
it seems to be a general issue these days
i think the core issue is that people are using aggregated 10 gigabit ethernet connections with people with high speed connections close by
and that seems to be more of an issue regionally, as long distance connections are 40g+
it may even be aggregated gigabit in smaller locations.

[17:19]

mnathani_

http://pastebin.com/byQUYHGx
thats from my residential 60mbit connection in Toronto

[17:31]

mercutio

wow you're so slow to japan
and nz
i wonder why
on teksavvy?

[17:32]

m0unds

[17:34]

mercutio

ok i find this more than a little disconcerting.
la is slower than it should be it seems
is a reoccuring theme

[17:35]

m0unds

haha

[17:35]

mercutio

i do wonder why mnathani was so slow to japan and nz though

[17:36]

m0unds

yea, i'm maxing out at 2.3MB/sec from that la test
could be horribly slow writes
yea, hitting 5.4MB/sec on my desktop
i forgot i was testing on my rpi :)

[17:38]

mercutio

it doesn't write
it redirects to > /dev/null

m0unds

well, an arm6 cpu is dog-ass slow
regardless

mercutio

yeah

brycec

mercutio: the Ethernet on the rpi is a USB connection to the host
so bus- and cpu-bound ;)

m0unds

and that ^

mercutio

yeh they suck

m0unds

it's fine for running crashplan
haha

[17:42]

brycec

I wouldn't say they suck, they just have their place

[17:42]

m0unds

yep
i have one that controls my fermenter and i have one running crashplan

[17:42]

brycec

(Find me another $35 computer that is comparably spec'd and has a non-usb and non-seial Ethernet controller :P)

[17:43]

mercutio

http://pastebin.com/4jcYkRjF
this is my vdsl connection

[17:43]

brycec

I have one for OOB/serial access to my core network, and one for Kodi. A good gamut of uses.

[17:43]

m0unds

rpi terminal server sounds good

mnathani_

I can send you an mtr

brycec

m0unds: it's powered by the USB port on my wifi AP :P

m0unds

i have some old-ass cisco 2516-rj
even better, hah
oh, i have one that does aircraft transponder monitoring via sdr dongle
forgot about that one

brycec

fun

[17:45]

m0unds

the dump1090 binary is a little crashy, but i just automate restarting it

[17:45]

mnathani_

mercutio: http://pastebin.com/nQR6qvNu

[17:47]

mercutio

yeah it's any2ix both ways
does that mean you're any2ix to la.meh.net.nz too?
maybe they're overloading their any2ix port

[17:48]

mnathani_

I redid the speed test http://pastebin.com/i8g3fz9x

[17:49]

mercutio

i'm probably going to have to upgrade japan's bandwidth :)

[17:50]

mnathani_

you mean japan your host
as opposed to the country right

[17:50]

mercutio

oh it's fine
yes
it's vultr tokyo server
108.713 GB of 200 GB (54%) (Bandwidth preferences)
it looks like it's fine because most of it's incoming

[17:50]

mnathani_

some hosts like liquidweb do free incoming

[17:52]

mercutio

vultr doesn't :/
but they onyl charge you whichever is highest
and outbound is way lower
on 42.468gb
it's a real pita increasing bandwidth on them
your vps shuts down for ages etc
cos you have to change plans

[17:52]

mnathani_

do other things change as well
like cpu cores and memory

[17:54]

mercutio

you get some more memory and more disk

[17:55]

mnathani_

when you enter the next tier

[17:55]

mercutio

yeah
but i don't use much disk
/dev/vda1 15G 3.1G 11G 23% /

[17:55]

mnathani_

do you have a slave setup for smokeping

mercutio

nope

mnathani_

or are they all individual

mercutio

i just stick it on multiple hsots

mnathani_

I guess its easy enough to compare / correleate events with the web interface

[18:02]

mercutio

yeah
i find it easier
but i want multiple views etc
so i can choose how it shows :)
so i'm trying arp's test download
and i'm finding it seems worse for performance
it seems to speed up slower

[18:03]

mnathani_

do you have 100 meg test files?

mercutio

nope
i can make one
i have 200mb
testfile.zip
it's urandom not a zip file

mnathani_

on each host?

mercutio

yeah that's everywhere
there's also 200k everywhere
i'm smokepinging with 200k
200k actually identifies issues often
but it has a huge latency bias
with 10mb the latency doesn't matter nearly as much
well at < 50 megabit
at > 200 megabit it does
they should all be on gigabit too i think
i also have it on cloudflare

http://weallsee.net/10m

but i don't think they like you doing that
cloudflare is served from that first ip
with light use it's prob ok

does someone else want to try hosting a 10mb file?

[18:20]

brycec

https://speedtest.cobryce.com/speedtest/random2500x2500.jpg ot
*it's 12MB but close enough
there's also 3000 and 3500
oh and 4000x4000 is 40MB
*30MB

[18:26]

mercutio

hmm it's slightly quicker
3598 vs 3407
so yeah probably just the larger file size
having a copy the same size woudl be better for testing comparitively i reckon

[18:27]

brycec

(and I have smaller files - 1000 1500 2000 2500)
(oh and 350 500 750)

[18:28]

mercutio

http://la.meh.net.nz/mini/speedetst/random2500x2500.jpg
i have that too :)
but not everywhere

[18:29]

brycec

good typo :p

[18:29]

mercutio

oh shit
s/detst/dtest/

[18:29]

BryceBot

<mercutio> http://la.meh.net.nz/mini/speedtest/random2500x2500.jpg

[18:30]

mercutio

oh yours is https
mine is faster for me now
oh cool you can take the s off

[18:30]

brycec

port 80 is there too
I didn't realize I linked to https

[18:31]

mercutio

yeah i'm finding mine slightly quicker
but not by much
i should check from uk
weird i'm getting way faster speeds from the uk to mine than yours
way faster than i was getting on 10mb before
like 1.6 to 2mb/sec to yours, and 4.4 to 5.4 to mien

[18:31]

BryceBot

That's what she said!!

[18:33]

brycec

brycec is also uploading to his VPS right now

[18:34]

mercutio

shouldn't matter?
oh it's a vps?
i thought it was dedicated from the mtr
but of cousre..
arp makes all vps's look dedicated :)

[18:34]

brycec

lol

[18:35]

mercutio

are you on 100 megabit?

[18:35]

brycec

Yes, I beleive so
*believe

[18:36]

BryceBot

That's what she said!!

[18:36]

mercutio

still, uk and nz are about same distance away

[18:37]

brycec

(there we go, flash speedtest fixed speedtest.cobryce.com)

[18:37]

mercutio

and it waas tiny difference from nz and huge from uk
i bet mine's out of date
yip
bah it wants you to login now
http://c.speedtest.net/mini/mini.zip

[18:37]

brycec

what?

mercutio

to download new mini.zip
my speedtest is out of date

brycec

It didn't ask me, but I guess I'm already logged in

mercutio

but chrome remembered my login
well it downloads that fast
http://la.meh.net.nz:24/mini/

brycec

from home, I get just better than 6mbps from yours

[18:40]

mercutio

weird now that was slower on mine than on yours for me
6 megabit?
i'm even getting slower upload

[18:40]

brycec

I find the mini test to be fairly unreliable, at least compared to what I get via other methods

mercutio

yeah it's lame

brycec

but it's a useful tool

mercutio

i prefer my 10mb test.
but what i really want is something that can test a whole lot of normal web sites etc.
but there's huge complexities in there
the problem with speedtest is it just tests close local servers normally
or you tell it to go to some server, adn you don't know which to go to
it doesn't really help getting 100 megabit locally if there are no web sites hosted locally
but how far do you go?
i often find ovh web sites slow
but i don't go to any of them regularly.
it's more i find a web site not loading and i trace to it
and it's ovh
and then it's back again sometime and still going slow
of course it's not necesasrily ovh's network
but like mnathani getting slow speeds to japan and new zealand may not really bother him at all

mnathani

I don't really need to reach nz or jp most of the time

mercutio

mnathani: exactly
i don't even need to reach japan really
but it's similar ping to california

mnathani

Did you see my second speedtest

mercutio

and it's another way to the US
so if there was some big outage between NZ and US I coudl bounce via there :)
oh they went u~p
i wonder why it was slow the first time.
so now sydney and la the slowest
i wonder what connection speed they have on any2ix
10 gigabit

http://pastebin.com/28uxqhES

maybe i should try upgrading kernel on la

[19:02]

ok slightly newer kernel

[19:13]

.................................. (idle for 2h46mn)

mnathani_

mercutio: 10 mb it too small for the speed test
here are my results with the 200 meg test file:

[21:59]

mercutio

mnathani_: why do you say taht?

[21:59]

mnathani_

[22:00]

mercutio

hmm la is still slow :(

[22:00]

mnathani_

connection takes a few seconds to reach max speed - by then the file download is already complete

[22:00]

mercutio

yeah but a lot of that's ramping up, the connection stability etc.
often for interactive stuff it's what matters
it's interesting how close nz and au are for speed
it does look like there's some issue there, but i can get fast to tehre
slower than last time
something's a little weird
it pauses around 7mb/sec for ages
but yeah for 200mb file in la from nz i get average speed of around 8 mb/seec, and current speed of aroudn 11.5mb/sec
that is faster than any of your test resutls.

[22:01]

mnathani_

I have another VPS I can test on
in Toronto

mercutio

oh yip

mnathani_

the tests from me were on residential connection

mercutio

i get 17mb/sec to sydney

mnathani_

are you talking megabits or megabytse?

mercutio

megabytes

mnathani_

lowercase means bits
:-)

mercutio

nah Mb means megabits
MB means megabytes
mb could be either :)
it shows as M in curl fwiw
the question remains why the slower speeds

brycec

17MB/s? Damn that's a nice fat pipe.

[22:12]

mercutio

uk is consistently giving 5200k/sec within 100k/sec
brycec: it's gigabit
that's testing from the top server on that list
i mean it's gigabit ethernet, not gpon or anything

[22:12]

brycec

heh
brycec is currently stressing his network and NAS pulling files @ 100+MB/s, wheee USB3 flash drive

[22:13]

mercutio

i can only do around 4.2megabytes/sec from home
heh i can do faster than that at home over network :/

[22:13]

brycec

My home network is only GbE :p

mercutio

yeah i ordered qdr infiniband cards

brycec

(averaging 108MB/s currently - disk IO bound)

mercutio

they're only $30 US each

brycec

Nice.

mercutio

it just means i have to get someone to resend them to me
i still don't know how much that'll cost

brycec

heh

[22:15]

mercutio

but ebay's global shipping program sucks
and the first seller wouldn't ship by usps
and the second seller doesn't ship outside the US

[22:15]

brycec

(Also IB wouldn't work well for me - can't very well run another cable across the apartment)

[22:16]

mercutio

heh
i'm still trying to figure out how the hell i'm going to cable my room
i want to get from one side of the room to the other
runinng cables along the ceiling is guly
ugly
running on the carpet with a rug over could damage them
atm i'm using ethernet and plugging and unplugging
but i'm thinking that i'll have a few on this side

[22:16]

brycec

I have a single cat6 strung over top the doorways from front to back of the apartment, it blends in more or less, but every time I see it is a reminder

[22:17]

mercutio

heh
well i was noticing that fibre is thinner than copper
but it's bright yellow
well the stuff i have is

[22:17]

brycec

(or orange, typically)

[22:18]

mercutio

this is smf
http://pastebin.com/6JyuxBEX
that's windows test to home on zfs
bloody samba is a cpu hog

[22:18]

brycec

Only a 100MB test file? Did that ever leave RAM?

[22:20]

mercutio

yeah
it's windows
it's disk cache sucks
if it was ram it'd be 2000+
i have 32gb of ram
it'd have to be huge to be bigger than ram
and it does seem consistent

[22:20]

brycec

lol

[22:21]

mercutio

well mostly
i'm using ssd's with write cache
err slc write cache
and linux will buffer some writes
but i care about network performance rather than disk performance for these tests :)
mnathani_: can you test downloading bryce's image?
there should be 5000x5000
i dunno what his cap is like heh
the strange thing about mnathani getting slower speed is that it's a any2ix route
and i'm getting fine any2ix performance
and japan and nz should be using any2ix too
and teksavvy only listened any2ix california rather than north/south

[22:21]

mnathani_

http://la.meh.net.nz/mini/speedtest/random2500x2500.jpg < that file?
or you mean from his box

[22:25]

mercutio

yeh but s/2500/5000/g
but yeah from his box
do both of them
see if it's similar

brycec

(it's an ARP VPS)

mnathani_

is the url handy

brycec

speedtest.cobryce.com/speedtest/random2500x2500.jpg

mercutio

oh there's 3500 but not 5000
oh you know what

[22:27]

brycec

up to 4000

[22:27]

mnathani_

are those actual images?

mercutio

if there's a transparent proxy it'll hit with bryce's link but not mine
yeah
they're just random pixels though

brycec

mnathani_: they'll render, sure... but it's static/random

mercutio

it's better than speedtest.net's upload test of zeros
but it's still dedupable
well you can also hack it to cache too
i was playing with it ;)
but it won't give you a result if it goes faster than 1 gigabit
i was just rewriting urls, it's the same mini.zip thingy for the real speed test servers too

mnathani_

I probably shouldnt test from 2 hosts simultaneously hitting brycec's vps

[22:33]

mercutio

haha
yeah he'll have 100 megabit
nothing has been extremely bad though, so it's diff from the packet loss issue

[22:35]

brycec

(and a portion of that is constantly in use :p)

[22:36]

mercutio

heh la.meh.net.nz has done like 1tb of data in the last 100 days
i thoguht that was quite a lot
in+out combined
oh i've lost it now because i updated the kernel

[22:36]

mnathani_

is that your dedicated box at ARP ?

mercutio

it was more like 900gb i think
it's a vm on dedicated box

mnathani_

so gigabit pipe

mercutio

yeh
but yeah there's a reason why i don't normally do 200mb tests ;)

mnathani_

atleast not automated ones
with smokeping

[22:39]

mercutio

well with smokeping it's 200k

[22:39]

brycec

brycec notices a few telltale spikes on today's ARP bandwidth graph ;)

mercutio

and it's pretty flat except sometimes gets better/worse
but stays good/bad
yeah i wonder if up_the_irons will see it on his graphs

mnathani_

my memory usage on windows desktop is at 88% and I cant seem to figure out whats using up 21 Gigs of Ram

brycec

it's relattively minor on mine, so almost certainly unnoticed.

mercutio

i think some people are doing huge amounts of bandwidth so maybe it'll fade into the background
mnathani_: haha

mnathani_

and I dont want to reboot

[22:41]

mercutio

i'm not even doing anything on windoms and it's using 5.7gb
two chorme windows, task manager, crystal disk mark, steam, a command prompt, skype, and foobar2000

[22:41]

brycec

hm I do have an arp metal box... I should set something up on it.

[22:42]

mercutio

chrome is runing a lot of processes for two windows

[22:42]

mnathani_

almost 90 chrome processes for me

mercutio

skype is using 174.1mx of ram
mb
how do you find out the number of chrome processes?

mnathani_

I counted
in the details

mercutio

god
i don't want to do that :)

mnathani_

probably a better way
using grep | wc -l
or something
if you have cygwin

mercutio

chrome://memory-redirect/
i don't have cygwin
i have synergy
i just use linux :)
chrome is using 1.3gb of private virtual memry

[22:44]

mnathani_

c:>tasklist | grep chrome | wc -l
85

[22:44]

mercutio

and 700mb of private memory
sj ;,ddk
oh sweet

[22:44]

mnathani_

when I installed git I chose the option to install unix tools in my path

[22:45]

mercutio

% cat tl | grep -i chrome | wc -l
16
i cut and paste it :)
i had tasklist
actually i piped it to a file, then used notepad on it and cut and paste it
cat tl | awk '{ print $5, $_}' | sort -n
will show you what's using the most ram
well "cat tl" can be tasklist for you

[22:46]

mnathani_

'awk' is not recognized as an internal or external command

mercutio

gawk?
mawk?
nawk?

mnathani_

c:>tasklist | gawk '{ print $5, $_}' | sort -n
-nThe system cannot find the file specified.

mercutio

haha shit
just cut and paste it to linux?
or bsd
same diff :)
hell even a mac will work
windows is so behind

mnathani_

there is probably a powershell equivalent

[22:49]

mercutio

i wonder what sort does
you maybe to output to a file and use the sort command
yeah sort doesn't take standard input
and you can tell it how much memory to use

[22:49]

mnathani_

task manager has a listing with memory
but its not accurate

[22:50]

mercutio

tasklist said 239,852 kb memory
and task manager said 171 mb i think it was
for skype

[22:51]

mnathani_

lol
turns out it was VLC
that had massive videos on pause

[22:51]

mercutio

don't you wish for "top"?
was it prebuffering the whole file or something?

[22:51]

mnathani_

it had a portion in memory
video was 48 gigs

[22:52]

mercutio

23gb though?

[22:52]

mnathani_

I was doing some video rendering earlier as well
process might just have been hanging in the background or something
I am running the curl speed tests on a Toronto VPS

[22:54]

mercutio

will be interesting to see how it compares

[22:55]

mnathani_

not sure what the max speed of this vps is

[22:56]

mercutio

you shall find out :)

mnathani_

wow

mercutio

comparative is intersting anyway

mnathani_

12.7 MB/s from LA

mercutio

i got more than that before
from nz
you'll hvae like 100 msec ping or something

mnathani_

you had gigabit though

mercutio

so do you that menas
means

http://pastebin.com/Bt9cXc6E

mnathani_

100 megabits / sec = 12.5 megabytes per second

[22:58]

mercutio

yaeh but 100 megabit physical ethernet never does 12.7mb/sec :)
it can sometimes do 11mb/sec

[22:59]

BryceBot

That's what she said!!

[22:59]

mnathani_

[23:00]

mercutio

hmm
that's actually lokoing really good

[23:00]

mnathani_

UK is a bit low

mercutio

tthe uk server doesn't haev good provider redundancy
it's just using level3
and some people don't have good connections to level3

mnathani_

but thats expected going over the pond

mercutio

not really
it's way closer than nz
japan is going surprisingly well
what speed do you get from that vps?

http://pastebin.com/XavbkEi5

mnathani_

its a customer vps
I never really tested

[23:03]

mercutio

oh that route is over any2ix too
it's any2ix/he.net

[23:03]

mnathani_

would you be surprised if route to jp from that vps woas out cogent

[23:04]

mercutio

it's what japan uses to get to it that matters
wow
the forward route is ntt->cogent too
it's going via seattle
what location is slc
oh is that salt lake city?
it's any2ix from nz too
but via northern side
hitting palo alto

[23:04]

mnathani_

[23:07]

mercutio

yeah the reverse route is basically the same
can you try to the nz server again?
i just upped the transmit window size

[23:08]

mnathani_

thats the one with the IP

mercutio

yeah

mnathani_

instead of hostname?

mercutio

it's emerald.meh.net.nz
if you want a name for it
bad habit :)
i know what it is :)

http://pastebin.com/HSw0G1bL

mnathani_

[23:11]

mercutio

slightly better
actualyl about 50% better
a bit less
but damn when did he.net get good? :)

[23:12]

mnathani_

just tested downlaoding to myself from the toronto VPS : http://pastebin.com/9cYizZzR

[23:14]

mercutio

nice
maybe you should bounce your connections via it haha
cable can often have inconsistent speeds
comcast, cox, verizon are all like that
they all have sub par routing
bulk/cheap
but from the little i read about teksavvy it sonuded like they wanted to be better than other people
so maybe they will peer more and more and upgrade their network etc.
but why some any2ix fast and some slow is bizzare
it doesn't seem like an arp issue
but it's hard to know for suure

[23:14]

mnathani_

I consistently max my connection while using usenet

mercutio

to a close server?

mnathani_

they use a pool and geo dns

mercutio

joh yip
oh yip
well any2ix is ages away from torronto :/

mnathani_

haven't run a trace to them

mercutio

but i can't for the life of me figure why it would be slower

[23:20]

mnathani_

http://pastebin.com/ddGVk8mY
thats one of the usenet servers
not sure if they do anycasting too

[23:21]

does appear to be an anycast address

[23:27]

mercutio

you suure?

[23:30]

mnathani_

pretty sure

[23:30]

mercutio

that's going to east coast for me it seems

[23:30]

mnathani_

I got low ping from ARP
and Toronto
and IL

[23:31]

mercutio

weird i get high ping from nz
sj to la to dc
connects to hwng in sj

[23:31]

mnathani_

how do I get mtr report to be 2 times as wide

mercutio

lots of things are high ping here though
like ebay

mnathani_

so hostnames are visible

mercutio

and amazon
amazon is around 230 msec
i dunno why it's so high
ebay lost hop it shows is phoenix at 150msec
but it seems to load slow
actually it's better than it used to be
i thought sites like that were meant to be heavily optimised thogh :)

mnathani_

try doing a dns lookup

mercutio

mnathani_

see if you get a different ip

mercutio

on what name?

mnathani_

news.newshosting.com

mercutio

nope
same location
news.newshosting.com is an alias for news.iad.newshosting.com.
i assume that is location