[TelstraClear] TelstraClear Cable connections appear to be rate shaped in one way or another

Forums › One New Zealand (including Vodafone, WxC, Farmside) › [TelstraClear] TelstraClear Cable connections appear to be rate shaped in one way or another

1 | ... | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12

2584 posts

Uber Geek

Retired Mod

Trusted

Lifetime subscriber

#526257 27-Sep-2011 15:37

I am sitting here struggling to get my head around how you are so convinced Telstra are not providing the service "promised". You can achieve the total bandwidth allowed. Maybe not on one FTP session to one server but you can get it. You are so much better off then the vast majority of broadband users out there that get a fraction of the total bandwidth they "should" be getting. Forget one TCP stream this is total bandwidth regardless of talking about one session or every thing they can throw at it.
Be thankful and move on with your life.

Media centre PC - Case Silverstone LC16M with 2 X 80mm AcoustiFan DustPROOF, MOBO Gigabyte MA785GT-UD3H, CPU AMD X2 240 under volted, RAM 4 Gig DDR3 1033, HDD 120Gig System/512Gig data, Tuners 2 X Hauppauge HVR-3000, 1 X HVR-2200, Video Palit GT 220, Sound Realtek 886A HD (onboard), Optical LiteOn DH-401S Blue-ray using TotalMedia Theatre Power Corsair VX Series, 450W ATX PSU OS Windows 7 x64

Ragnor

8223 posts

Uber Geek

Trusted

#526264 27-Sep-2011 15:48

jnawk:

XNet can do it. What is different in their setup?

Telstraclear seem to have some kind of upstream issue at play, hopefully they can get to the bottom it but to address your current expectations....

Dedicated bandwidth/transit (both domestic and international) is very expensive. Residential ISP's do not buy bandwidth at a 1:1 ratio with users, in fact no way near 1:1, probably more like 400:1.

Residential Internet is able to be priced at consumer level prices because it's a shared network best effort service.

The only part of your connection to the world where you are not sharing bandwidth (contended bandwidth) with other users is the copper from your house to the cabinet/exchange.

Backhaul from the cabinet/exchange to the ISP handover point = shared between users
ISP domestic bandwidth to various places = shared between users
ISP international bandwidth to various place = shared between users.

Once it gets past the ISP's network they have little control, once it gets past their upstream providers network they have no control.

Depending on customer demographics, time of day and the ratio of users to bandwidth the ISP operates on their various links.... you normally won't run into congestion but it is by no means guaranteed for the price you are paying.

No ISP will guarantee that you will get downloads from even national locations nevermind international at line rate on a residential ADSL connection (normally you will though).

It's the nature of the beast, if your expectations are otherwise they are unrealistic.

jnawk

176 posts

Master Geek

#526267 27-Sep-2011 15:50

Fact they are looking into it speaks volumes as to whether or not they are delivering what they "promised".

Fact also is other ISPs can deliver total bandwidth to a single TCP session (TCP is an end to end protocol, so all the bits in the middle shouldn't care if its one session or many).

If I had a 4mbit connection, I'd be happy with the speed I see. If I had DSL, I'd be well aware that a rubbish line speed is the gamble you take. But I don't have DSL, I don't have a rubbish line rate. I'm paying (one way or another) for 15mbit/sec. I don't want or need that kind of speed, but I have no slower options (at least, none that don't cost the same or more, or aren't dial up). I figure while I'm paying for 15mbit/sec, I may as well receive it. What packets I choose to send are my business, the ISP claims to not shape. If that were the case, my packets would flow at 15mbit/sec, down one connection or many. And yes, it'd fluctuate with usage, I accept that. But every time I get 3.5mbit/sec, I can get more if I up the number of connections. And it is directly proportional.

jnawk

176 posts

Master Geek

#526270 27-Sep-2011 15:53

Ragnor:
jnawk:?

XNet can do it. What is different in their setup?

Telstraclear seem to have some kind of upstream issue at play, hopefully they can get to the bottom it but to address your current expectations....

Dedicated bandwidth/transit (both domestic and international) is very expensive. Residential ISP's do not buy bandwidth at a 1:1 ratio with users, in fact no way near 1:1, probably more like 400:1.

Residential?Internet?is able to be priced at consumer level prices because it's a shared network best effort service.

The only part of your connection to the world where you are not sharing bandwidth (contended bandwidth) with other users is the copper from your house to the cabinet/exchange.

Backhaul from the cabinet/exchange to the ISP handover point = shared between users
ISP domestic bandwidth to various places = shared between users
ISP international bandwidth to various place = shared between users.

Once it gets past the ISP's network they have little control, once it gets past their upstream providers network they have no control.

Depending on customer demographics, time of day and the ratio of users to bandwidth the ISP operates on their various links.... you normally won't run into congestion but it is by no means?guaranteed?for the price you are paying.

No ISP will guarantee?that you will get downloads from even national locations nevermind international at line rate on a residential ADSL?connection (normally you will though).

It's the nature of the beast, if your expectations are otherwise they are unrealistic.

I'm on cable, so its shared all the way. But that isn't really the point.

I'm not asking for dedicated bandwidth. I'm only asking for no shaping of individual connections. If I can get 15mbit in aggregate (I can - on demand), there is no reason why I can't get it down a single connection.

Ragnor

8223 posts

Uber Geek

Trusted

#526272 27-Sep-2011 15:53

They should definitely look into a clear problem like this, if they can fix it I'm sure they will but if they can't you are out of luck.

You are paying for up to 15Mbit/2Mbit, residential broadband (either cable or adsl) has no guarantee of performance or reliability.

There are plenty of hops in the tracerts to your servers outside of their control, ie: not their network and not their upstream providers network where the single thread limiting could be occurring. Also the the return path could be the issue and there's little they can do about that.

It's not clear cut, hopefully if it's in their network they can find and fix it.

jnawk

176 posts

Master Geek

#526283 27-Sep-2011 16:04

Ragnor: There are plenty of hops in the tracerts outside of their control, not their network and not their upstream provider where the single thread limiting could be?occurring.?

I could accept that if it were an issue with a handful of hosts. But the same thing happens to a server on the other side of the planet, and the hops after their immediate upstream are totally different.

And it can't be happening in the part of the network that is common to both traces (ie, the last few hops), because then it would affect both connections (mine & the XNet one).

Beccara

1469 posts

Uber Geek

ID Verified

#526291 27-Sep-2011 16:25

jnawk:
Ragnor: There are plenty of hops in the tracerts outside of their control, not their network and not their upstream provider where the single thread limiting could be?occurring.?

I could accept that if it were an issue with a handful of hosts. But the same thing happens to a server on the other side of the planet, and the hops after their immediate upstream are totally different.

And it can't be happening in the part of the network that is common to both traces (ie, the last few hops), because then it would affect both connections (mine & the XNet one).

There is a hop within your providers network that is different between ISP's that isn't at their edge

Most problems are the result of previous solutions...

All comment's I make are my own personal opinion and do not in any way, shape or form reflect the views of current or former employers unless specifically stated

Sharesies

Trade NZ and US shares and funds with Sharesies (affiliate link).

jpollock

600 posts

Ultimate Geek

Trusted

#526298 27-Sep-2011 16:33

Nety: I am sitting here struggling to get my head around how you are so convinced Telstra are not providing the service "promised". You can achieve the total bandwidth allowed. Maybe not on one FTP session to one server but you can get it. You are so much better off then the vast majority of broadband users out there that get a fraction of the total bandwidth they "should" be getting. Forget one TCP stream this is total bandwidth regardless of talking about one session or every thing they can throw at it.
Be thankful and move on with your life.

This is all based on my understanding, I'll be doing my own experiments to Amazon's EC2 servers in the next couple of weeks. :)

The problem is that I think jnawk can saturate his 15mbps pipe with 4-5 sockets to the same server, but he can't saturate it with 1 socket to that server, not even close. This affects services such as VPNs (working from home), video streaming, and other high bandwidth TCP services. It doesn't affect bittorrent, since that consists of large numbers of individual TCP connections.

Now, there is a known issue with base TCP when the ping time (RTT) to the server gets above 30ms (see sbiddle's first response, and my followup explanation), it starts to limit it because the base TCP receive window isn't big enough. However, both endpoints support an extension which allows for larger window sizes making it go faster - this is an agreement between endpoints, the middle tiers shouldn't care, nor should they modify this functionality. I'd like to say that this isn't a dodgy extension either, Vista, OSX, and Linux have all supported it for ages.

Additionally, I have heard that Telstra Clear while testing saw UDP traffic saturate the pipe (90mbps) between the server and the client, indicating that this is a TCP-only problem. A second ISP isn't seeing this problem with TCP.

I noticed this problem previously, but I figured that TCL was shaping the connections. Since they've come on here and said they don't shape, that changes the discussion.

So, what we have here is a problem where the endpoints agree to use an extension, the extension is being transmitted, we should be seeing fast traffic but aren't. Switching to UDP results in full traffic, so this isn't a contention issue, nor is it a problem with there not being enough to go around - the traffic is getting through. This is someone playing games with TCP traffic and only TCP traffic.

I'm fine with them needing to rate limit to support improve contention, but do it to my IP, not individual sockets.

Jason

---
http://blog.jason.pollock.ca

Ragnor

8223 posts

Uber Geek

Trusted

#526306 27-Sep-2011 16:40

So an interesting test would be to see if you are still limited to 3-4Mbit on a single tcp download from a server in Australia outside of Telstrclear and Reach's network?

Try a large download from one of these?
http://games.on.net/filelist.php
http://www.3fl.net.au/downloads/

Also would be interesting to see if you get the same result from say Singapore, Hong Kong etc.

jnawk

176 posts

Master Geek

#526400 27-Sep-2011 18:33

Ragnor: So an interesting test would be to see if you are still limited to 3-4Mbit on a single tcp download from a server in Australia outside of Telstrclear and Reach's network??

Try a large download from one of these?
http://games.on.net/filelist.php
http://www.3fl.net.au/downloads/?

Also would be?interesting?to see?if you get the same result from say Singapore, Hong Kong etc.

I got 1.2MB/sec (9.6mbit/sec) off of that second site, RTT is 100msec. However, I advise caution interpreting this result - there's a transparent HTTP proxy in the way, remember, and it's a damn site closer than 100msec!

sbiddle

30853 posts

Uber Geek

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#526429 27-Sep-2011 19:20

Ragnor: So an interesting test would be to see if you are still limited to 3-4Mbit on a single tcp download from a server in Australia outside of Telstrclear and Reach's network?

Try a large download from one of these?
http://games.on.net/filelist.php
http://www.3fl.net.au/downloads/

Also would be interesting to see if you get the same result from say Singapore, Hong Kong etc.

Results will be significantly faster because the RTT is lower.

PenultimateHop

637 posts

Ultimate Geek

Trusted

#526430 27-Sep-2011 19:24

jnawk: Best effort also doesn't mean traffic be shaped. Which has to be the case - we've proven each endpoint and all the hops along the way have capacity, and that there is no droppage occurring. Its just TCP that suffers.

No you haven't. And you didn't answer my other questions either.

I agree with the others that you're pushing quite to the limit to say this is a fault or outside of the service spec.

PenultimateHop

637 posts

Ultimate Geek

Trusted

#526431 27-Sep-2011 19:28

Also if you'd like a thorough debugging of this consider providing pcaps of each transfer from both client and server end, for both client types.

jnawk

176 posts

Master Geek

#526436 27-Sep-2011 19:40

PenultimateHop:
jnawk: Best effort also doesn't mean traffic be shaped. Which has to be the case - we've proven each endpoint and all the hops along the way have capacity, and that there is no droppage occurring. Its just TCP that suffers.

No you haven't. And you didn't answer my other questions either.

I agree with the others that you're pushing quite to the limit to say this is a fault or outside of the service spec.

If the capacity was not there, I'd not be able to get 15mbit/sec. That I can get 15mbit/sec (I can) clearly indicates the capacity is there.

sbiddle

30853 posts

Uber Geek

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#526446 27-Sep-2011 20:21

jnawk:
PenultimateHop:
jnawk: Best effort also doesn't mean traffic be shaped. Which has to be the case - we've proven each endpoint and all the hops along the way have capacity, and that there is no droppage occurring. Its just TCP that suffers.

No you haven't. And you didn't answer my other questions either.

I agree with the others that you're pushing quite to the limit to say this is a fault or outside of the service spec.

If the capacity was not there, I'd not be able to get 15mbit/sec. That I can get 15mbit/sec (I can) clearly indicates the capacity is there.

You of course keep making the assumption that something is broken. What if it's not?

1 | ... | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12