大象传媒

Mind your MBps and Mbps's there! Mbps usually refers to megabits per second, while MBps refers to mega bytes per second - a big difference (1 byte = 8 bits so 8x - although usually taken as 10 for reasons too obscure to go into).

Very nice memories seeing that 1200/75 modem - it was great for loading Prestel pages (1K each) - the days before the Internet had pictures :).

That got me thinking - it would be interesting to have a graph of content size against bandwidth. Connections are getting faster, but digital things are getting bigger.

The digital camera I was using last week was a 14 Mega pixel model - a single image from that wouldn't have been able to fit on to a collection of hard drives from 25 years ago.

It's the same scenario as building more roads - you'll just find more cars to fill them!

Were the seeders of the torrent in the UK/Europe? That can make a big difference to the speed that is seen

BenjaminEllis - I don't actually see a single misuse of MBps/Mbps. If you're referring to the upload speed mentioned, I think Virgin media may actually be that slow upstream.

1200/75? That's luxury!
My first online experiments were using a 300 baud acoustic coupler on BT Gold sometime in the early 80s.
:-)

Thanks for the updates, Darren - it's very interesting to read a 'real' end-user's account.

I too downloaded the halo wars demo when it first came out through BT and it took me at least an hour to download it completely via wireless on a 3-4mb line so if the 50mb is as good as virgin says then it someone I will seriously consider

I am sooo jealous, Here in Malaysia, our fastest speed is only 4Mbps...

Really, those tests are useless in real life.
The suggestions and tests are completely obscured. upload and download to an ftp site? thats reasonable, but these are only on virgins internal network servers. but the other tests? theres no technological reasoning for these, you havent even pointed this out in your article.

You want to test your connection? You want to test your connection to the backbone and beyond. Once your at the backbone, your into the internet, your not affected by virgins servers.
youre on the internet!

Firstly, if you want a real download rate, download off a real download site (e.g rapidshare) with a download manager. you will get a true download speed with that.
Yes, its going to create many connections to the server, but it WILL saturate your connection. giving you a real download rate.

running a speed test from speedtest.net is also a good idea. it runs a test from servers located on backbones closest to you. giving you a realistic upload and download rate with a ping.

the rest is down to other outside servers.
no other tests are relevant.
therefore youre tests are completely pointless.

I cant believe this:
"but as soon as I threw in some background downloads of large files then the picture started to stutter on one of the machines."
this is a joke seriously?! this is caused because TCP is unbias against any connection, it is unbias against all traffic. when your streaming something and downloading something, tcp doesent care, its going to download everything evenly. therefore your saturating your connection with the downloads and streaming.
any connection will do this.

Testing your connection via wireless is also ridiculous. Your internet speed is different on wireless, it fluctuates? ITS WIRELESS! your internal network affects this, the quality of your equipment and distance from the router will completely affect the test.

seriously, this article has really surprised me. I would have thought that the 大象传媒 could at least employ someone with some tech knowledge, with capabilitys to explain it in plain English.

im at uni, ive got a 100mbs connection to the janet backbone. i download and upload at 10mbs. i get a ping to servers in maidenhead on the backbone at 10ms, and bbc.co.uk at 11ms. After that, its not my ISPs fault, im on a backbone, im on the internet. Run some real tests along the right lines.
i could go on. but its boring and im being obnoxious.

@william0will: I believe the purpose of some of the tests is to see if the Virgin 50Mbps service really does deliver 50 Mbps to the door, and the only way to test that is to use Virgins own servers. As soon as you move out onto the internet, then you are no longer testing your connection to Virgin but Virgins connection to the backbone, the backbone itself, and the remote servers connection. By running a set of different tests you isolate and baseline the performance of the different segments and can then analyse where any bottlenecks may lie. It's a basic performance analysis technique and one you should learn if you work in the IT world.

While TCP itself doesn't care what is in a packet, it doesn't evenly distribute the packets it just works on a first in best dressed basis. However, the use of QoS on most routers these days does prioritise traffic at least within your own network, and probaly witihin most ISP's as well. Network neutrality is important on teh backbone, but I'm quite happy to prioritise my Skype calls and iPlayer traffic over my news feed and e-mail server.

If you had read the entire series of blogs, then you would see that most of your points have been raised before, especially the wireless one, which led to Darren making several improvements to his wireless network that produced near saturation through-put.

@william0will

"Testing your connection via wireless is also ridiculous. Your internet speed is different on wireless, it fluctuates? ITS WIRELESS! your internal network affects this, the quality of your equipment and distance from the router will completely affect the test."

This point has already been established in the blog posts and some tests have been done using both wireless and wired connections to compare them. A large number of people use wireless these days and therefore it seems right that this is the standard environment for the testing. You point out yourself that the tests need to resemble real life usage and wireless fits that bill well.

"running a speed test from speedtest.net is also a good idea. it runs a test from servers located on backbones closest to you. giving you a realistic upload and download rate with a ping."

You may not have read Darren's comment about the online speed tests he used therefore I have quoted it below.

From the 'day 5' blog:-
"I took readings of the connection speed from a number of different measurement tools across the week and found there was a slight variance in the speed"

Maybe you should read all the information first and you may not feel the need to be so obnoxious (your word, not mine)

and from the introductory diary blog:-
"Of course the first thing I did was test the speed. Virgin says that only two one website - thinkbroadband.com and speedtest.net had tests that could successfully gauge the speeds available to its users.

I used both tests - on both a wired and wireless connection"

Like i said the FTP test was reasonable. but you really dont have any clue what these ftp servers are capable of. I was saying that there should have been more practical real world tests, and tech tests using appropriate tools.
I suggested rapidshare, because files are hosted on high bandwidth distributed servers capable of 100mbs+ download speeds. This is well known.
Yes, i was going to mention QoS but it wasnt appropriate. HTTP downloads, and iplayer both run over port 80, and cannot be classified on a standard home router.
TCP algorithms give even priority to http connections, this is why this has occurred.
This is due to the home network, not the connection.

Some of the ways the connection was tested from the suggestions were rubbish i think.
He didnt want to download anything off the pirate bay, so he used a legal website with few seeds and peers, that isnt a good test.
he could have just downloaded a linux iso and got some decent stats from that.

My point was that some of the speed tests from user suggestions were not really carried out with the accuracy, knowledge in ways i would expect. The suggestions werent that great, and this wasnt raised by the author. he didnt point out why the videos stuttered when downloading a large file.
There are many tools and ways of testing connections more reliably with more accurate results.

@Darren
In the past I've downloaded a few things I've missed on iPlayer from a UK based TV site, and with the high number of UK seeders, the speed is phenomenal, maxing out the 20Mb Virgin line for the duration of the download (all 10 mins of it!). I'm not sure what BT client and settings you were using, but I would be willing to bet that the defaults won't even begin to make full use of the line and number of connections it can support.


@william0will
There are always 2 benchmarks that any system has to meet - the numerical and the subjective.
While I would agree that most readers of these blog posts are probably interested in the purely numerical stats (i.e. is Virgin delivering the 50Mb line that they've promised, which it seems they are), actually demonstrating a real-world interpretation of the 'geek-speak' and what the connection can do for mere mortals who have no idea what QoS, TCP or packets are is far more relevant.
Yes, perhaps you or I would solve the problem by employing more sophisticated traffic management hardware and software (as I suggested in response to the Day Four post), but as you said you really need to be knowledgeable in the field to even come up with the suggestion.

I presently work on the autopilot system for one of the next-gen of aircraft. As a similar example, the pilot doesn't necessarily need to know the full technical details of how we resolve an issue they have, they just need to know what practical things resolving that issue enables them to do with the aircraft.

The blog shows that the advertised headline speed is fairly irrelevant to the actual experience of an average customer, even when the provider can actually deliver such speeds reliably within the confines of their own network.

The limitations, especially in the newly designed cable broadband networks, are not at the access layer of the network (the wire from the local network node to the home). They are out in the internet world beyond.

It just goes to show that messages about headline speeds can only disappoint. Maybe it's time for Virgin Media to change their marketing, rather than fiddling further with their technology.

You said: "I think it is safe to say that we have made some great leaps in terms of internet connection speeds over the last 25 years."

I'm not sure sure the leaps are great everywhere, as BT say my maximum expected Broadband speed is 125Kbps (and that the maximum, which of course it never runs at.

I have declined their offer to sign up,

I don't know how long I'll be waiting for up to 8Mbps broadband, never-mind the 50/100 a lot of people and the 大象传媒 seem to be talking about a lot recently.

"Each of the five files I simultaneously downloaded was coming down at almost 1.0MB per second."

ie 8Mb per second - ie a total of 40 Mb per second - good to see that Virgin is offering what it can't deliver - even from its own FTP servers...

@richardwhiuk

There will be connection overheads that mean that you'll never see 50Mbps numerically, but the connection was probably maxed out. A BT download from just Virgin users would perhaps provide a better test, but practically impossible to set up!

I have a question...

Why can my 拢280 PS3 download at my full downstream speed indictated on the router while my 拢500 laptop struggles to download faster than 120kbps on dedicated fast servers?

I think technology in general has been left in the hands of capitalism for far too long. As long as businesses are allowed to sit back and upgrade nothing in their search for pure profit only, the human race will never know how to invent something as apparently simple as the toilet that prevents splash-back let alone 50MB broadband across the nation.

Raw speed is not so importatant - but enough speed is.
There is a level of performance below which you are pushing rocks uphill but once you break through there are always other limiting factors.
Rather like buying a car once you have a car with adequate performance upgrading to a Ferarri may not increase your speed (due to the car in front).

Once you can go uphill at 70 miles an hour without changing gear its enough. lees than 75 horse power is a problem but much over 150 horse power is not making so much difference, due to other bottlenecks a long the way
5 Mbps is a realistic minimum

I have (as non UK resident)for some long time 2 fast connections - 25Mps (Docsis cable)at my home in the Netherlands and 60Mbps (DSL) in my office in Germany.
While both peform adequately I canot say I notice a big difference.

I have now the option to upgrade the cable to 125Mbps fibre link for a samll fee - but it is a question if this is interesting except as a topic of conversation.

Here 8Mbps is the lowest entry level subscription and costs typically less than half of UK price
I pay only 20 UKpounds per month for 25Mbps connection including VAT
Random testing show I have always better than 20Mbps.

It is a building requirement here for the last 5 years that all new houses have a fibre connection (for 120Mbps) also for TV and Phone.
For 30years all new houses have cable TV connection - that can also supply Internet and telephony
Plus we have a telephone line that can supply also Interent and digtal TV

Less than 2% of the population have satelite TV.

So we have in built competition - - when customers have the option to switch supplier at an instant - my experience is that performance improves by magnitudes and you hear nothing about investment recovery, customer serice lines are answered in sconds and not hours and the tone is quite different


It looks to me to be very simple