[Snap] UFB Poor Performance

Forums › 2degrees (including Slingshot, Orcon, Flip, Stuff Fibre, MyRepublic, 2talk and Vocus) › [Snap] UFB Poor Performance

1 | ... | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | ... | 19

1418 posts

Uber Geek

ID Verified

#712167 5-Nov-2012 12:04

mercutio:
hamish225:
Ragnor:
sbiddle: The challenge for Chorus / LFC's and ISP's is actually engineering the CIR - in effect allocating 2.5Mbps of guaranteed traffic to users when this bandwidth will sit dormant for most of the time.

Apart from the odd issue EUBA has worked relatively well with no per user dimensioning.

It doesn't matter because probably no one is going to use the voip port on the ONT nor the CIR tagged traffic for consumer plans.

Pretty much every ISP is just using EUBA0, not bothering with vlan tagging and just sending voip over the best effort internet bandwidth then doing some QoS/management/engineering of traffic inside the ISP network to smooth things out.

This is probably going to continue for UFB

Having CIR's for special tagged traffic on the consumer bitstream offerings seem like a retarded waste of time to me.

Business plans sure.

maybe they should make regualr http traffic have the 2.5Mb/s cir, so that the guy down the road with the dedicated torrent machine doesn't slow down the tech savvyy grandma who likes to browse trademe quickly?

2.5megabit/sec isn't quick.

still better than nothing i guess, why even have it then, what's the point?

*Insert big spe*dtest result here*

ubergeeknz

3344 posts

Uber Geek

Trusted

Vocus

#712208 5-Nov-2012 13:05

Non CIR traffic is treated "fairly" so someone leaching won't get "all teh bandwidths". If saturation is reached in any part of the network, packets will be queued and/or discarded at random from all connections.

Also, the network is provisioned so there is "at least" the CIR available per connection, and it's very unlikely that tagged traffic will ever saturate that.

You can't simply prioritise "all traffic" - it defeats the whole purpose. And defining more levels (ie. AFxx for HTTP traffic) would just add a lot of complexity for IMO little benefit.

quakeguy

111 posts

Master Geek

Trusted

#713392 7-Nov-2012 10:23

freitasm: I mean someone told me there's a suspicion why some UFB connections are not performing as well as intended. Apparently not all ISPs are aware and some are investigating. Let's see what they find.

I've just caught something nasty on EUBA and I'm proceeding to investigate UFB with the same idea in mind. Surprising. I'm actually an advocate for Chorus' network normally, as overall I've seen very good performance in recent years (really since the 7302 upgrade/Ethernet backhaul project).

“I do not think there is any thrill that can go through the human heart like that felt by the inventor as he sees some creation of the brain unfolding to success... Such emotions make a man forget food, sleep, friends, love, everything.” - Nikola Tesla

Disclaimer: Views expressed in my posts do not necessarily reflect those views of my employer.

SneakerPimps

104 posts

Master Geek

#716126 13-Nov-2012 01:19

Dump... Snap has been quiet about this.

bisr

227 posts

Master Geek

#716134 13-Nov-2012 06:35

SneakerPimps: Asus RTN-16 with Tomato firmware.

I'll give the direct connection a go, but I'm pretty sure the router I have is capable of the throughput.

That RT-N16 is awful. I have one too, it's a piece of sh*t. I doubt however that it's the router's fault at all because I get 18Mbps easy on ADSL2 going through it.

Lorenceo

904 posts

Ultimate Geek

Trusted

#716247 13-Nov-2012 11:11

<3 my N16. It can easily handle 100mb both up and down. Would be nice if it had 5ghz wireless too, but that's not a big issue.

SneakerPimps

104 posts

Master Geek

#716251 13-Nov-2012 11:22

I've deployed about a dozen of the RT-N16 with Tomato and it's solid, never had an issue with it. bisr you would be first person I've heard to speak bad of it :/.

It's not a router issue because what I'm experiencing is mainly international speed. Plus I've tried another router now, and it performs the same.

//EDIT - ACTUALLY, I take that back. I've just tested this again now and it seems to be fixed!!!!

Equinox IT

Cloud spending continues to surge globally, but most organisations haven’t made the changes necessary to maximise the value and cost-efficiency benefits of their cloud investments. Download the whitepaper From Overspend to Advantage now.

quakeguy

111 posts

Master Geek

Trusted

#716260 13-Nov-2012 11:32

SneakerPimps, can you please try the following:

http://test1.snap.net.nz/test-file-100m

Let me know the speed this downloads, both single-threaded and multi-threaded.

Can you also please PM me your username, so I can set up a graph of latency to you, also to track packet loss.

Thanks,
Tim

Disclaimer: Views expressed in my posts do not necessarily reflect those views of my employer.

SneakerPimps

104 posts

Master Geek

#716263 13-Nov-2012 11:37

1 Thread = ~2MB/s
4 Threads = ~3.1MB/s

Lorenceo

904 posts

Ultimate Geek

Trusted

#716297 13-Nov-2012 12:27

That speedtest.net result is possibly inaccurate. Snap seem to have started caching HTTP, including speedtest.net, so you can no longer use it to test international speeds. :(
See here: http://www.geekzone.co.nz/forums.asp?forumid=90&topicid=111499

SneakerPimps

104 posts

Master Geek

#716316 13-Nov-2012 12:44

hhmmm interesting.

I'll be able to tell more later, but just quickly checking, most links on this thread to test download files have almost double in speed. (from 150KB to ~300KB/sec).

Improvement yes, but at least it's a start.

mercutio

1392 posts

Uber Geek

#716321 13-Nov-2012 12:50

Lorenceo: That speedtest.net result is possibly inaccurate. Snap seem to have started caching HTTP, including speedtest.net, so you can no longer use it to test international speeds. :(
See here: http://www.geekzone.co.nz/forums.asp?forumid=90&topicid=111499

speedtest.net uses unique requests for each "random image" that it uses to download, and so even if things force-cache the content, it'll just take up lots of cache space.

so you'll have a request like:

http://speed.snap.net.nz/speedtest/random1500x1500.jpg?

then a random number after the ? .. you can request the image yourself and see it just looks like random pixels.

but with the http download, you should be able to accomplish the same thing with that 100mb speed test in fremont (california) by appending ?129321039812903812 or some other random number to it, and increment it every time you do a new test.

uknzguy

19 posts

Geek

#716568 13-Nov-2012 17:42

quakeguy: SneakerPimps, can you please try the following:

http://test1.snap.net.nz/test-file-100m

Let me know the speed this downloads, both single-threaded and multi-threaded.

Can you also please PM me your username, so I can set up a graph of latency to you, also to track packet loss.

Thanks,
Tim

Snap UFB 30M customer here. My downloads are pretty poor too -

root@nucleus:/tmp# time wget http://test1.snap.net.nz/test-file-100m
--2012-11-13 17:47:37-- http://test1.snap.net.nz/test-file-100m
Resolving test1.snap.net.nz... 64.71.180.170
Connecting to test1.snap.net.nz|64.71.180.170|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 104857600 (100M) [text/plain]
Saving to: `test-file-100m'

100%[=================================================>] 104,857,600 1.48M/s in 63s

2012-11-13 17:48:40 (1.60 MB/s) - `test-file-100m' saved [104857600/104857600]

real 1m2.942s
user 0m0.088s
sys 0m1.324s

sbiddle

30853 posts

Uber Geek

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#716583 13-Nov-2012 18:17

I'd still really love somebody who's experiencing slow speeds and has the technical knowledge to run some tests using the high priority CIR on their connection rather than just their EIR.

I'm just curious if the connection is actually meeting the specs for the UFB product which is only around the CIR, not the EIR which at the end of the day is all best effort traffic and congestion could be occuring at many points along the way.

Lorenceo

904 posts

Ultimate Geek

Trusted

#716642 13-Nov-2012 19:34

mercutio: speedtest.net uses unique requests for each "random image" that it uses to download, and so even if things force-cache the content, it'll just take up lots of cache space.

so you'll have a request like:

http://speed.snap.net.nz/speedtest/random1500x1500.jpg?

then a random number after the ? .. you can request the image yourself and see it just looks like random pixels.

but with the http download, you should be able to accomplish the same thing with that 100mb speed test in fremont (california) by appending ?129321039812903812 or some other random number to it, and increment it every time you do a new test.

I just tried this, with a bunch of different number strings. The MD5 of the files that it downloads are the same, and they all came in at line speed. Even when supposedly coming from Africa.

1 | ... | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | ... | 19