UFB 100/50 speedtests?

Forums › 2degrees (including Slingshot, Orcon, Flip, Stuff Fibre, 2talk and Vocus) › UFB 100/50 speedtests?

1 | 2 | 3

435 posts

Ultimate Geek

Trusted

#680010 1-Sep-2012 10:06

Here's a download of that file on my Snap 100/50 UFB connection

MightyApe

Affiliate link

Affiliate link: You will find anything you want at MightyApe.

Zeon

3861 posts

Uber Geek

Trusted

#680022 1-Sep-2012 10:37

Maybe Orcon are already having CIR contention issues?

Speedtest 2019-10-14

freitasm

BDFL - Memuneh
74136 posts

Uber Geek

Administrator

ID Verified

Trusted

Geekzone

Lifetime subscriber

#680035 1-Sep-2012 11:04

Publius: @Domoth did a test on my orcon 100/50 UFB downloading http://nz.archive.ubuntu.com/ubuntu-releases/12.04/ubuntu-12.04.1-desktop-amd64.iso and getting ~50mbit/sec down. I've never been able to get my connection over ~50mbit/sec

@Zeon router is orcon-supplied cisco SRP521W which should be adequate. Thanks for the confirmation that games.orcon.net.nz is able to saturate over 100mbit/sec.

I'd still like to hear from anyone getting >50mbit/sec over orcon ufb, or anyone's ufb for that matter.

I have a Cisco SRP521w here (TelstraClear HFC network) and I've noticed it max at 80 Mbps. Also notice it has a 100 Mbps WAN port which should really be a 1 Gbps port.

Having said that I have tested a Linksys E4200 with WAN/LAN gigabit ports and it didn't perform much faster than that - limited by CPU/memory I think not being able to move packets at the speed it needs to achieve full line speed.

Also make sure you are doing tests over ethernet, not over wireless.

freitasm on Keybase | My technology disclosure

mercutio

1387 posts

Uber Geek

#680037 1-Sep-2012 11:14

freitasm:
Publius: @Domoth did a test on my orcon 100/50 UFB downloading http://nz.archive.ubuntu.com/ubuntu-releases/12.04/ubuntu-12.04.1-desktop-amd64.iso and getting ~50mbit/sec down. I've never been able to get my connection over ~50mbit/sec

@Zeon router is orcon-supplied cisco SRP521W which should be adequate.? Thanks for the confirmation that games.orcon.net.nz is able to saturate over 100mbit/sec.

I'd still like to hear from anyone getting >50mbit/sec over orcon ufb, or anyone's ufb for that matter.

I have a Cisco SRP521w here (TelstraClear HFC network) and I've noticed it max at 80 Mbps. Also notice it has a 100 Mbps WAN port which should really be a 1 Gbps port.

Having said that I have tested a Linksys E4200 with WAN/LAN gigabit ports and it didn't perform much faster than that - limited by CPU/memory I think not being able to move packets at the speed it needs to achieve full line speed.

Also make sure you are doing tests over ethernet, not over wireless.

I noticed the other day a router available in NZ that actually does hardware NAT (like the google fibre router does) so should be suitable for fast connections while not using lots of power up.

http://www.tp-link.com/en/products/details/?categoryid=2166&model=TL-WDR4300

It's rated at 800 megabit wan to lan throughput which should be plenty for future proofing.

You want a router that can do considerably more than 100 megabit if you don't want to have issues with smaller packets, bidirectional use, being used as a wireless access point at the same time, and to not slow/congest the connection down.

80 megabit seems pretty low. I can get that with 2.4 ghz wireless.

And as you said all ports should be gigabit now days, the latency difference is minimal until you start congesting the connection up, but overall it should keep jitter down, and will be able to operate at a full 100 megabit peak, rather than around 90 megabit or so.

mercutio

1387 posts

Uber Geek

#680039 1-Sep-2012 11:17

I was just reading about NBN somewhere, and it was saying that providers over there get charged $20/megabit for transit to the fibre connections. That's going to lead to some heavy national congestion.

It actually sounded like UFB was more sensible than NBN in some ways. Although NBN had a better packet loss target than UFB for voip. I dunno if I'm overexpecting, but the cir class for UFB sounds appropriate for bulk traffic to me, it's classed as 0.1% packet loss 1msec jitter <= 5 msec ping. 0.1% packet loss is enough to hamper 100 megabit throughput nationally.

Found a link about the $20/mbit and that'd be AUD. http://www.tektel.com.au/TekTel%20Report%20-%20NBN%20Pricing%20Explained.pdf

Anyway, GPON isn't meant to be a "always high speed" connection. If you really want a fast connection you're meant to get the P2P fibre, otherwise it's pretty much best effort, and your experience is likely to be variable.

There can be bottlenecks at the router, at the fibre provider (chorus, enable, etc), at your service provider's entry point from fibre network (orcon etc), at your service provider's exit point to other people's networks, at other people's network point entry point, at other people's local network, at other people's san/storage.

If you use conventional sata disks, in raid 10, and you do lots of simultaneous uploads of multiple files you are likely to bottleneck easily with non-cachable large downloads. It's a bit better for downloads as writes can be kept sequential and bunched up as disk latency increases, but for reads you need intelligent read-ahead which can suddenly start to fail as load gets too high.

In normal usage of the internet these days you're not likely to burst anywhere near 100 megabit. A considerable amount of web sites take longer to generate the page and to initiate connections to them, then to be throttled by your connection.

Just try doing a 10 mb download, and even if you get 100% of your connection at the end, the average will be significantly lower. It's even more so with 200k or less, where web objects generally are.

And lots of games like say Starcraft, use patch systems that use seperate files and seperate server locations within the same patch download, and thus you get things like going 2 megabytes/sec at the start, then at 90% it seems to start from 0, and go up to 200k/sec or something.

freitasm

BDFL - Memuneh
74136 posts

Uber Geek

Administrator

ID Verified

Trusted

Geekzone

Lifetime subscriber

#680047 1-Sep-2012 11:37

mercutio: I was just reading about NBN somewhere, and it was saying that providers over there get charged $20/megabit for transit to the fibre connections. That's going to lead to some heavy national congestion.

It actually sounded like UFB was more sensible than NBN in some ways. Although NBN had a better packet loss target than UFB for voip. I dunno if I'm overexpecting, but the cir class for UFB sounds appropriate for bulk traffic to me, it's classed as 0.1% packet loss 1msec jitter <= 5 msec ping. 0.1% packet loss is enough to hamper 100 megabit throughput nationally.

Reading iStart magazine?

freitasm on Keybase | My technology disclosure

mercutio

1387 posts

Uber Geek

#680052 1-Sep-2012 11:57

freitasm:
mercutio: I was just reading about NBN somewhere, and it was saying that providers over there get charged $20/megabit for transit to the fibre connections. That's going to lead to some heavy national congestion.

It actually sounded like UFB was more sensible than NBN in some ways. Although NBN had a better packet loss target than UFB for voip. I dunno if I'm overexpecting, but the cir class for UFB sounds appropriate for bulk traffic to me, it's classed as 0.1% packet loss 1msec jitter <= 5 msec ping. 0.1% packet loss is enough to hamper 100 megabit throughput nationally.

Reading iStart magazine?

Was actually reading: http://dts.net.nz/blog/nbn-vs-ufb-head-to-head-comparison/

Hadn't heard of istart.

sbiddle

30853 posts

Uber Geek

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#680067 1-Sep-2012 12:40

99% of people have no understanding what the CIR means though anyway - they certainly don't realise the CIR aspect of UFB is only high priority tagged traffic, not non tagged best effort traffic. The logistics of managing this is also massive - in effect Chorus and other LFC's need to dedicate a guaranteed 2.5Mbps to every property. This also means aggregation switches need to be scalled appropiately, and ISP's need to have massive amounts of dedicated bandwidth to support this CIR which may sit there unused.

It also creates problems - start running VoIP over this CIR and like HSNS packets are dropped rather than queued when you exceed it. This means if you're deploying services such as video calling you need to be fully aware of the exact bandwidth requirements because the minute you exceed them things implode.

mercutio

1387 posts

Uber Geek

#680076 1-Sep-2012 13:27

sbiddle: 99% of people have no understanding what the CIR means though anyway - they certainly don't realise the CIR aspect of UFB is only high priority tagged traffic, not non tagged best effort traffic. The logistics of managing this is also massive - in effect Chorus and other LFC's need to dedicate a guaranteed 2.5Mbps to every property. This also means aggregation switches need to be scalled appropiately, and ISP's need to have massive amounts of dedicated bandwidth to support this CIR which may sit there unused.

It also creates problems - start running VoIP over this CIR and like HSNS packets are dropped rather than queued when you exceed it. This means if you're deploying services such as video calling you need to be fully aware of the exact bandwidth requirements because the minute you exceed them things implode.

in essence it's unsuitable for high quality video calling.

what i don't understand about the best effort class - is that it says <= 2% packet loss, so does that mean on a 100 megabit connection you should be able to do 98 megabit of received traffic?

it's one thing to have 2% packet loss at 100 megabit - it's quite another to have 2% packet loss at 10 megabit. it'll severely effect connection quality at that level.

also as far as i understand it isp's don't have to dedicate cir. Only the service provider (chorus etc) ?

And I very much doubt that aggregation of international traffic will be anything like 2.5 megabit international per user.

I know my individual usage is under 2.5 megabit average. But my "cir" traffic would be significantly under that. VOIP is what, 90kbit/sec/channel with g711 voice over sip. And on DSL without I seem to manage fine on straight euba0.

I don't know what's better - dropping, or queueing in the CIR, but I'd guess if you're just over dropping is better as long as it's dispersed and not bursty. TelstraClear Cable for instance seems to have burst loss - if you send 30 packets in a burst, it may get 1 through 16, then drop 17 through 24, then give you 25, then drop through 28, then give you 29, 30...

If you look at average packet loss it's 33% but not every 3rd packet. Although, packet loss can be a complicated and confusing thing - like doing a ping may show no loss, but you may have loss inside an individual stream.

I don't know what solutions "international" companies are doing for VOIP, but sometimes I've rung companies who provide a NZ number that obviously is going over VOIP, and it's terrible - and having a CIR locally or not isn't going to change that. A few examples of such are: Blizzard (going to US I think), LG (not sure where), and some generic local access number pass through thing.

I'm assuming codec could be an issue there, with Blizzard there was also latency and echo (sounds like you're in a cave).

The other thing is - are these VOIP services going to be reliable for home users. Over the years I've seen countless router issues, packet loss issues, congestion, outages. Recently it's been getting worse, not better.

There used to be issues with ISP's having outages, you ring them up, it gives you a recorded message saying that due to congestion your call cannot be put through. Now you ring up, it answers, then drops the call.

At least Slingshot and Orcon have recently had "isp issues" that took out their phone system. They're also two of the providers doing UFB trials right now.

It's one thing to have 10 msec jitter, it's quite another thing to have no phone service for hours on end.

I think the safe solution right now is to have plain old telephone backup, but I am concerned that there isn't enough warning about the inherent risks of relying on a VOIP service for people who are used to telephones being reliable and stable. And I think SLA's are more significant than CIR's in that respect.

I for one would much prefer 2.5 megabit CIR over whole connection without split VOIP. If I really had to rely on VOIP for anything critical, then I'd want two connections through two separate providers. Right now I have a hack of a prefix to choose between 2talk and wxc for outgoing calls, but once I found that both of them went down at once -- all traffic to APE was being dropped, and oh, hangon, they're both routing through APE. Google was fine. Maybe I'd need an Australian provider for fallback.

I personally don't really care about video calling that much, but it's the main advantage I see of IP calling for normal users, the second being it makes it easy to do two calls at once. Recent phones have 720p front cameras, I don't know if they have realtime h264 encoding, but from my understanding that is about 1.5 megabit/sec, so you'd need to either have some kind of culling of frames, or to run other non CIR bandwidth for a second phone call.

In the US they have free wifi calling on cellphones that goes off your normal cellphone plan. I for one prefer to use a cellphone with wifi calling then a cordless phone. My cordless phone doesn't fit in my pocket - and even if it did I'd have my cellphone in my pocket too, and it seems kind of redundant.

I suppose the real magic will really be new super high definition conference calling. - 2560x1440, and up, or even multiple streams, so that meetings can be had from a distance and capture multiple people and be blown up and look good. Screen technology is going up, and you can already get even cellphones that do 1080p, but in the next step 10 megabit would be more appropriate, and yet users aren't going to want to pay for a 10 megabit cir. (you'd probably be fine with 720p upload sitting in front of the camera)

So I think in a way I wouldn't hold it against any provider who decides that the CIR is a kind of legacy thing and can basically be ignored.

I, myself, am actually more interested in G722 wideband. But as it stands now because landlines and cellphones aren't using it, and there's other compression algorithams in use it's not really of much benefit.

Domoth

16 posts

Geek

#680092 1-Sep-2012 14:38

I did another test this morning with the Ubuntu iso and got the following results:

4.x MB/s is about as fast as its been for me since getting fibre.

Its cool to hear I am not an isolated case here. I would be keen to try new hardware but need the voice and fear trying to deal with Orcon on non standard hardware. I was hoping going to fibre meant we no longer had excuses from ISPs about wiring and distance from cabinet / exchange. Sadly it looks like we have a new set of problems now. I am still optimistic these might be teething problems though.

Ill post how my service call goes. They hopefully will be getting back to me this week.

mercutio

1387 posts

Uber Geek

#680093 1-Sep-2012 14:46

Domoth: I did another test this morning with the Ubuntu iso and got the following results:

4.x MB/s is about as fast as its been for me since getting fibre.

Its cool to hear I am not an isolated case here. I would be keen to try new hardware but need the voice and fear trying to deal with Orcon on non standard hardware. I was hoping going to fibre meant we no longer had excuses from ISPs about wiring and?distance?from cabinet /?exchange. Sadly it looks like we have a new set of problems now. I am still optimistic these might be teething problems though.

Ill post how my service call goes. They hopefully will be getting back to me this week.

have you tried downloading multiple files in parallel?

You will probably need to ring them if they say they're going to return phone call.

Domoth

16 posts

Geek

#680103 1-Sep-2012 15:14

have you tried downloading multiple files in parallel? You will probably need to ring them if they say they're going to return phone call.

Ok, tried 3 at once with a much better result:

I learnt to chase Orcon about call backs a long time ago

kyhwana2

2539 posts

Uber Geek

#680145 1-Sep-2012 17:43

Domoth:

have you tried downloading multiple files in parallel? You will probably need to ring them if they say they're going to return phone call.

Ok, tried 3 at once with a much better result:

I learnt to chase Orcon about call backs a long time ago

See, the above looks fine to me. You're getting 7.5MB/s / 60mbit/s and 2/3rds from the same source. Don't know what kind of connection games.orcon.net.nz has to where and what the usage was at the point you downloaded it..

Try a few more connections at around 3-4am..

mercutio

1387 posts

Uber Geek

#680150 1-Sep-2012 17:52

kyhwana2:
Domoth:

have you tried downloading multiple files in parallel? You will probably need to ring them if they say they're going to return phone call.

Ok, tried 3 at once with a much better result:

I?learnt?to?chase?Orcon about call backs a long time ago?

See, the above looks fine to me. You're getting 7.5MB/s / 60mbit/s and 2/3rds from the same source. Don't know what kind of connection games.orcon.net.nz has to where and what the usage was at the point you downloaded it..

Try a few more connections at around 3-4am..

# curl -O http://games.orcon.net.nz/files/ubuntu/ubuntu-12.04-server-i386.iso
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
100 673M 100 673M 0 0 50.3M 0 0:00:13 0:00:13 --:--:-- 51.5M

that's 50 megabytes/sec

that last link he linked to was in europe though. 2megabytes/sec to europe is pretty good.

Publius

276 posts

Ultimate Geek

#681236 4-Sep-2012 11:37

I finally found a command line program which supports multiple connections to download 1 file.
$ axel -n 4 -a http://games.orcon.net.nz/files/ubuntu/ubuntu-12.04-desktop-amd64.iso

1 connections downloads at ~32mbit/sec
4 connections downloads at ~58mbit/sec
12 connections downloads at ~70mbit/sec

Which makes me wonder what is going on.

traceroute to games.orcon.net.nz (219.88.241.85), 30 hops max, 60 byte packets
1 xxxxx (60.234.x.x) 0.519 ms 0.436 ms 0.394 ms
2 ufb.bng1.nct.orcon.net.nz (121.99.252.1) 5.543 ms 5.729 ms 5.894 ms
3 xe-2-2-0.cre1.nct.orcon.net.nz (121.98.9.141) 4.325 ms 4.275 ms 4.226 ms
4 games.orcon.net.nz (219.88.241.85) 4.640 ms 4.602 ms 4.773 ms

Could someone from Orcon comment? Speedtests that others have posted also on UFB 100mbit from other isps show they can get 80% of their connection speed without having to resort to extreme measures like this.

Whats more is i'm a business connection so if some sort of traffic shaping is happening then i'm not going to be impressed.

1 | 2 | 3