[TelstraClear] How Fast is Telstra Clear to the San Francisco or Los Angeles, USA?

Forums › One New Zealand (including Vodafone, WxC, Farmside) › [TelstraClear] How Fast is Telstra Clear to the San Francisco or Los Angeles, USA?

1 | 2

michaeln

238 posts

Master Geek

Trusted

#65949 2-Apr-2007 10:36

Mac OS X has a default RWIN of 64kB (cf WinXP 16). That means that the maximum throughput from a West Coast server will be about 3.5Mbps and for an East cost one, 2.3Mbps) (cf 870kbps and 570kbps for Windows). You can increase the RWIN for Mac OS X by doing the following in Terminal.app sudo sysctl -w net.inet.tcp.recvspace=237500 (if you are comfortable with that). The Apple Supplied BroadBand Tuner will do this for you automatically http://www.apple.com/support/downloads/broadbandtuner10.html.

And as Maurice said: make sure you distinguish between kB and kb.Note, some web pages will come in very much faster than is 'theoretically' possible.That's becuase they are served locally through Akamai (the MS and Apple software updates), or in the cache.

[Moderator edit (MF): hyperlinked]

inane

216 posts

Master Geek

Trusted

#65956 2-Apr-2007 11:11

just to add to the unit size conversations,

the maximum theoretical throughput of a 10megabit cable connection is

1.2megabytes per second, iirc.

bhv

12 posts

Geek

#65959 2-Apr-2007 11:25

I'm measuring speed using testing services like Speakeasy.

http://www.speakeasy.net/speedtest/

Here is a typical result for a test to the west coast of the USA. Other speed tests give about the same results.

Last Result:

Download Speed: 799 kbps (99.9 KB/sec transfer rate)

Upload Speed: 750 kbps (93.8 KB/sec transfer rate)

Yes I am talking about bits (eight of them per Byte). Right now I'm downloading Firefox at about 56 KB/s =446 kb/s.

Obviously Telstra Clear is not responsible for the performance of the entire net, but they should get you to the backbone at the advertised speed.

Performance after that point obviously will vary. Right now it seems that I need to look at adjsting the size of RWIN.

Networks designed to go really long distances (think Earth to the Moon) obviously have to be configured much differently than shorter range networks.

If indeed the intrinsic latency to the US means that you will NEVER achieve speads even remotely close to advertised speeds to the US (without some serious network tweaking) then Telstra Clear should publish that fact (along with how to do the tweaks) and save themselves a lot of grief from annoyed customers.

In the mean time, my most sincere thanks to those here who have helped me understand the problem better. Perhaps one of the network gurus from this group will write a FAQ about how to tweak your computer for maximum performance over the long fiberoptic networks on which our data must travel.

bhv

lchiu7

6470 posts

Uber Geek

Trusted

#65972 2-Apr-2007 12:54

I am not on the Mac platform but recently when downloading a bittorrent file (legal to be sure so well served) I achieved 650kB/sec on the 10Mbs plan. This from mainly US peers and is about the best I have achieved from the US. As a matter of interest they were videos of the fan fiction versions of shows in the Star Trek universe. While they downloaded fairly fast, the quality of the content is another discussion :-(

Larry

Staying in Wellington. Check out my AirBnB in the Wellington CBD. https://www.airbnb.co.nz/h/wellycbd PM me and mention GZ to get a 15% discount and no AirBnB charges.

bhv

12 posts

Geek

#65973 2-Apr-2007 13:08

Hi Michaeln,

Thanks for the tip on Broadband Tuner from Apple.

I downloaded it, think I installed it (I'm a serious Mac newbie), and here's the result. The test numbers are about the same, but my

"User Experience" with sites like http://www.amazon.com is much better.

Perhaps it's the placebo effect? [smile]

The software is only needs to update 4 numbers or so, so it's pretty tiny and I didn't see any place to "run" the software from

after installing it. Perhaps the installation is the only thing necessary.

I can't remotely figure out why the download speed numbers could stay the same while my perception is that pages load much more quickly, but

that is the case.

Pages DO seem to load much more quickly. I'm not sure if it's the placebo effect or not, but it does seem to be progress.

Thanks,

BHV

michaeln

238 posts

Master Geek

Trusted

#66011 2-Apr-2007 19:20

A site like Amazon has lots of images. The TCP 'slow start' algorithm means that any increase in speed, even a small one, tends to be magnified. If you have 20 images, and each of them only takes a little less time to load, the time taken to load the whole page will be noticeably quicker.

tknz

182 posts

Master Geek

#68999 1-May-2007 00:39

The speed is definately consistent, what you are getting is 10/mb per second of BANDWIDTH, e.g the pipe coming into your house, I use rapidshare downloading which is top speed uploader, and i consistently get 1100ish KB/s Downstream from international servers, the problem is that it requires multiple connections to a server you won't get one single ten megabit persecond pipe to one server. because not only that they would have to be able to upload that the WHOLE way there will be a number of routers in between the traffic will be going all sorts of different places before it gets to you it will try find the fastest connection per route. so if one is limited to say 1000kb/s from one server and you make a second connection it may go a different route.

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.

michaeln

238 posts

Master Geek

Trusted

#69022 1-May-2007 09:55

Tuikapo, BGP (Border Gateway Protocol), which is the 'glue' that binds the Internet together, does not find you the fastest path. It finds the 'shortest' path (for some values of 'shortest' that may not always be quite what you might expect.). If a path gets congested, BGP does not try to find a less congested one. If there are multiple paths, then it MAY share them. When topology changes, for instance if a path goes down, BGP finds another path. Also, if there ARE multiple paths, each individual packet in a connection will choose between them: you will not have one session go one way and a second session go the other way.

Fraktul

836 posts

Ultimate Geek

Trusted

#69120 1-May-2007 19:32

The effect of multiple streams acheiving higher throughput to the same server is a result of the TCP RWIN being set to low for the path latency as previous discussed.

Michaeln: Not quite true, if a router has two routes to the same destination and is set to per session balancing then it will send all of one session down down a specific route as opposed to balancing it per packet. Refer Ciscos CEF balancing. This is handy in some applications to prevent out of order delivery.