Serious degraded performance outside of Orcon Network

Forums › 2degrees (including Slingshot, Orcon, Flip, Stuff Fibre, MyRepublic, 2talk and Vocus) › Serious degraded performance outside of Orcon Network

1 | ... | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | ... | 24

409 posts

Ultimate Geek

#1305552 15-May-2015 06:39

as long as you have throughput on EIR traffic ufb is working as designed, regardless of the speed rate.

Satch

1985 posts

Uber Geek

#1305763 15-May-2015 13:04

Peppery: I understand that UFB is 'best effort' bandwidth so I'm trying to be reasonable, but how close to the 'advertised' speeds am I meant to get, even locally? We pay for 100/50, however even on local Speedtests (to Orcon) we never seem to top more than 70mbit. Tests done over LAN with minimal other users on the network...

I think this is a very fair comment to be honest. I know that the people who understand this stuff a whole lot more than I do keep saying it is only best effort and if you wanted guaranteeed speed that you'll have to take out a more expensive business contract with the ISP, but common sense also has to prevail. The best effort argument stacks up only if you experience occasional degradation of service, and that is reasonable in my mind. But to consistently have speeds degradation of 35% plus must fall outside that catch-all.

I wonder what your ISP would say if you were to say that paying your bill in full was a best effort?

BlackHand

131 posts

Master Geek

#1305827 15-May-2015 13:56

My problem is even though its a best effort service for as long as I've had my 100/50 fibre connection (more than a year) I use to get at least 90/40 anytime of the day, since mid-April its degraded to 60-70/20.

PassiveFever81

367 posts

Ultimate Geek

#1305830 15-May-2015 14:02

@blackhand
Similar situation here, for the past couple years I got 27/9 consistantly, since April, 23/9 down to 15 at night.

Still havent heard anything back from Orcon regarding the testing I have been doing...

Jase2985

13462 posts

Uber Geek

ID Verified

Lifetime subscriber

#1305872 15-May-2015 14:49

http://www.geekzone.co.nz/forums.asp?forumid=40&topicid=171348&page_no=6#1305599

have a read of that post, it may be applicable in this case

FlameBeard

344 posts

Ultimate Geek

Trusted

#1305919 15-May-2015 15:52

Jase2985: http://www.geekzone.co.nz/forums.asp?forumid=40&topicid=171348&page_no=6#1305599

have a read of that post, it may be applicable in this case

an interesting read for sure. The thread applies really only to say that VF has issues with the Wellington Region and capacity.

4th gen i7 Haswell 4770k, G.SKILL RipjawsX 16GB (4x4 Gb) DDR3 2400MHz, x1 GTS 460, Intel 180Gb 530 Series SSD, x1 Seagate 1Tb HDD, x1 Seagate 2Tb HDD, Modular 850w PSU, R.O.G. Maximus VII Formula mobo, Cooler Master Storm Trooper Chassis, Cooler Master V8 CPU cooler

"Five exclamation marks, the sure sign of an insane mind." - Terry Pratchett

No longer work for Orcon

kiwikiwi

455 posts

Ultimate Geek

#1305941 15-May-2015 16:17

FlameBeard:
Jase2985: http://www.geekzone.co.nz/forums.asp?forumid=40&topicid=171348&page_no=6#1305599

have a read of that post, it may be applicable in this case

an interesting read for sure. The thread applies really only to say that VF has issues with the Wellington Region and capacity.

An exert from both the thread and the PDF file.

Advertised download or upload speeds should be realistic and consistently achievable by consumers. They should not reflect theoretically achievable or maximum speeds as it is unlikely that customers will be able to get these speeds regularly given factors such as the customer’s distance from the exchange, the number of people using the service and the quality of the customer’s wiring. Broadband suppliers should have a reasonable expectation of being able to provide services that they advertise. Representations that suppliers can provide speeds “up to” a theoretical maximum may breach the Fair Trading Act if that maximum speed is not realistically achievable by consumers.

I think this actually applies here. I'm gonna use the word "previously" a lot. Previously I was able to achieve 94mbps down and 57mbps up theoretically from speedtest. In reality I was easily achieving 11.2MB(89.6mbps)/s and able to set up a direct multi thread FTP upload and max out at 5.7MB(45.6mbps)/s. So previously yes I was getting exactly what I paid for wtihin reason and up to. That was until 5 weeks ago. So where does this actually put the ISP I am with(Orcon here) now?

Right now it's breaking about 2 of those bolded parts. My download and upload speeds are no where near what they are and haven't been for 5 weeks straight. It's also no where consistent either.

So in my opinion I'd say yes they are breaching.

You can also follow me on twitter here @kiwifortw I do twitch streams every now and then at twitch.tv/kiwiforthewin :)

HTTP 404 Sarcasm not found.

Quic Broadband

Move to New Zealand's best fibre broadband service (affiliate link). Free setup code: R587125ERQ6VE. Note that to use Quic Broadband you must be comfortable with configuring your own router.

Jase2985

13462 posts

Uber Geek

ID Verified

Lifetime subscriber

#1305964 15-May-2015 16:50

FlameBeard:
Jase2985: http://www.geekzone.co.nz/forums.asp?forumid=40&topicid=171348&page_no=6#1305599

have a read of that post, it may be applicable in this case

an interesting read for sure. The thread applies really only to say that VF has issues with the Wellington Region and capacity.

yes the thread does but the info in that post and my one 2 down apply to any ISP where there is performance issues

andybrrr

30 posts

Geek

#1306217 16-May-2015 10:08

East Auckland 100/20 UFB at 10 AM on a Saturday

Just checked to Spark now and got

And so checked Sydney again and ..

In the space of 3 minutes .. I don't even know lol edit: Sorry about formatting, I can't even work out how to do a line break right now

dwl

371 posts

Ultimate Geek

#1306225 16-May-2015 10:40

kiwikiwi: So in my opinion I'd say yes they are breaching.

I know this thread is titled "performance outside the Orcon network" but it isn't clear to me whether you are ever getting full rates (e.g. 94 Mbps for 100Mbps service) to an Orcon server. I believe it is reasonable to expect the ISP to achieve the advertised speed most of the time within their network but once outside there are many more variables at work.

If outside the network it gets tricky. Say you are using a Spark test server and not getting very good results. While Spark probably isn't deliberately limiting traffic to Orcon what a great way to make another ISP look bad and grab some customers. Provided Spark customers can source traffic from Orcon at high rates they probably wouldn't complain.

Note that there is a big difference between what they are contracted to provide (remember the old ADSL CIR of about 26kbps per user was all that was promised) and what should be available for most of the time as best efforts.

skylight01

54 posts

Master Geek

#1306245 16-May-2015 12:30

Can confirm the speeds are a problem within orcons network, They escalated my complaint and had me do various speed tests within the network including ftp transfers of 500mb files to and from there servers.
My speeds have been sitting at around 60-70% of what I used to achieve, they mentioned in the conversation many times that they did not intentionally cap the traffic.

Their final say was they contacted chorus and told me 60-70% is within the contract.
I aksed if they just oversigned and congested their networks there was much avoidance in answering haha. I asked if this is getting fixed if not i'll go to another isp which can provide closer to advertised speeds. Reply was that yes they will fix this as they want to retain their reputation as the number one fibre provider.

So i guess orcon better provide some dates as there will be some customers that wont accept this issue

dwl

371 posts

Ultimate Geek

#1306295 16-May-2015 14:01

Minh: Their final say was they contacted chorus and told me 60-70% is within the contract.

This raises a couple of important issues.

If they are suggesting that the congestion is occuring within the Chorus network then it could be the local segment or the number of users on a handover. If local then I think we have an issue as a country because this could start affecting other ISPs as well. If at specific handover locations (aggregated users in a city) that should hopefully be within Orcon control to segment the users.

The more concerning aspect is if a speed of 60-70% for local traffic becomes considered as acceptable, the exact characteristics of the loss and delay that results in these rates becomes important. As has been said multiple times, TCP will reduce rates more for higher delay sources than low delay sources for a given loss profile. Once rates less than full speed within the Orcon network are considered acceptable (and I'm not suggesting Orcon is saying that yet, they are just pointing out lack of contractural commitment) I think it becomes inevitable that quite poor performance could result from higher delay sources outside their network.

I am sure that Orcon are looking at this seriously and won't be wanting customers to leave. Hope some resolution is available soon.

kiwikiwi

455 posts

Ultimate Geek

#1306351 16-May-2015 16:34

dwl:
kiwikiwi: So in my opinion I'd say yes they are breaching.

I know this thread is titled "performance outside the Orcon network" but it isn't clear to me whether you are ever getting full rates (e.g. 94 Mbps for 100Mbps service) to an Orcon server.

Originally it was purely outside yes but it's also turned into an internal issue too. Even my own speedtests to their actual speedtest server, pings to their game servers and website have gone down hill.

I believe it is reasonable to expect the ISP to achieve the advertised speed most of the time within their network but once outside there are many more variables at work.

I'd accept that if my data was going to America due to distance, but no. Most game servers and most of my traffic does unfortunately go outside of New Zealand. I am able to max out my connection to Australia. During peak times it gets really messy and I start seeing latency+speedtests go down hill. I'm not just testing spark no I do test speedtest.orcon.net.nz which is inside their network apparently and it gets just as bad as a speedtest to Spark or Aussie.

You can also follow me on twitter here @kiwifortw I do twitch streams every now and then at twitch.tv/kiwiforthewin :)

HTTP 404 Sarcasm not found.

dwl

371 posts

Ultimate Geek

#1306435 16-May-2015 22:01

This is geekzone so I have had a play with a couple of Linux boxes using netem (network emulation, part of the kernel) to try and see the effect of delay and loss on TCP transfers. I started with local delay and some loss to achieve a slight reduction in speeds (as is being reported) and compared to what happens with higher delay for the same loss. The shaping to limit to 100 Mbps was simple tail drop. The RTT reported is minimum with buffering delay additional:

Netem loss 0.0%, delay 10ms; 341.6439 MB / 30.13 sec = 95.1031 Mbps 94 retrans 11.93 msRTT
Netem loss 0.5%, delay 5ms; 306.3213 MB / 30.19 sec = 85.1123 Mbps 1449 retrans 7.02 msRTT
Netem loss 0.5%, delay 10ms; 272.0731 MB / 30.68 sec = 74.3842 Mbps 1048 retrans 11.98 msRTT
Netem loss 0.5%, delay 50ms; 107.8982 MB / 30.34 sec = 29.8365 Mbps 330 retrans 51.94 msRTT
Netem loss 0.5%, delay 180ms; 44.1839 MB / 35.90 sec = 10.3240 Mbps 133 retrans 182.01 msRTT
Netem loss 0.0%, delay 180ms; 259.8716 MB / 30.36 sec = 71.7970 Mbps 0 retrans 181.93 msRTT

The loss value of 0.5% was chosen to give a noticeable reduction in speed with low delay but resulting in speeds that might be considered adequate. As the emulated network delay increased the TCP throughput dropped off significantly. The last result with 180 ms delay with 0% loss was added to show the window size was not an issue - it was slow to ramp up and had achieved a sustained 95 Mbps after 15 seconds.

This highlights for me the importance of ISPs maintaining high performance within their networks if we are to get the full benefit of UFB. For higher delay sources there may also be additional loss elsewhere and it is easy to see how the throughput performance could collapse.

kiwikiwi

455 posts

Ultimate Geek

#1306475 17-May-2015 08:37

And now my phone service is dead again.

Joy.

You can also follow me on twitter here @kiwifortw I do twitch streams every now and then at twitch.tv/kiwiforthewin :)

HTTP 404 Sarcasm not found.

1 | ... | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | ... | 24