We've found that telstra throttles all kinds of things on the way into and out of Australia. So much so that a simple openvpn tunnel between our rack there and our rack in NJ (USA) makes just about everything better.
The Australian broadband market is highly segmented according to speed. Until quite recently Telstra Bigpond was offering a 256Kbit connection labelled 'ADSL broadband' with 512Kbit and 1.5Mbit connections priced significantly higher. The idea of an 8Mbit or 20Mbit connection would be almost science fiction to their traditional customer base.
Now that FTTH is coming into play in some areas and Bigpond's market share is dropping, higher speeds are becoming more affordable but there are still a lot of people on those lower speed plans. I suspect this is what is skewing the graph labelled 'ISP', more than any selective throttling by the ISP (though I certainly wouldn't put that past them either).
The comparison to other ISPs is not entirely fair either because their customers tend to be clustered in areas where the ISPs have their own hardware in Telstra's exchanges and can therefore offer higher speeds and attractive pricing. Elsewhere in the country, Bigpond and lower speeds are both much more prevalent.
4.29 Mbps average, in melbourne, significantly higher than all the other rates (including my ISP, iinet). It bursts up to 8 Mbps.
It's probably because I'm very close to the exchange (I checked before I moved).
FWIW, it has made no impact on my life satisfaction whatsoever. People rave about highspeed internet, esp in Japan; it sounds cool, but the applications aren't there yet (not for me, anyway). I'd appreciate lower latency, though, for games.
If you look at future applications , there isn't a big benefit from higher bandwidth.
For video , 720p takes 2mbit(youtube's 720p) , and the difference between 720p to 1080p isn't very big.if we go to 3D , it might be doubled , but with smart compression , might be less.
For video conferencing , the compression is much more effective because of the small frame to frame changes, so i think you can get 1080p at 2mbit up + 2mbit down.if i recall right , in the right setting , 1080p is needed to give you the illusion of being in the same room.
For bit-torrent,to make streaming possible(i think streaming is the most natural way of video consumption), probably what is needed is higher uploads ,and configuration such that uploading doesn't slow the download side , not a higher download bandwidth.
For video game streaming(onlive.com) , what's important is the latency . i believe the speed is good enough.
The other future application of fast bandwidth is using cloud storage as your HDD. not a very attractive application if you ask me. dropbox is good enough.
So i don't see anything in the horizon requiring very high bandwidth internet for the home. 10-15 down / 4 up would be more than enough for any average family.
1080p is actually significantly better than 720p on large screen sizes at standard computing distances (or if you have 20/20 eyesight and sit at standard TV distances from a ~50" or larger screen). I'd also expect at least 4mbit for streaming 720p feature movies in H.264.
Very high bandwidth links are also useful for lowering the delay involved in bulk transfers like uploading videos, and a higher link speed (even if the total bandwidth is throttled) results in shorter individual packet transmission times, and thus lower interactive latency.
I use a 28" monitor with a resolution of 1920x1200. I've watched the same video in 1080p and 720p, and the 1080p is appreciably sharper and more detailed, even accounting for the fact that the 720 version is being scaled to the display's native resolution. It is 2.25 times as many pixels, so there's over twice as much detail possible. It's kind of fun to read the signs and computer monitors in the background of scenes (for example, on a 1080p version of The Island, you can see a character's hilariously disturbing criminal record on a monitor in the background).
As for latency sensitive applications, going from a 2mbit link to a 20mbit link reduces the round trip transmit time (excluding interpacket delays required by common physical links) of a 256-byte packet from ~2ms to ~0.2ms. If both endpoints migrate from 2mbit to 20mbit, then ~3.6ms of latency can be eliminated. The benefits are larger for larger packet sizes, so larger packets can be used (and therefore increased throughput) while maintaining low latency.
I think you missed a current application that needs a stack of bandwidth — network boot or install of operating systems. MacBook Air's already do a wireless network install from another machine and ChromeOS will be delivered solely via the internet. You can dismiss the prospect in the same way you've dismissed cloud storage but given trends in physical media for music and video, I don't think a decade will pass before doing an OS upgrade or install that requires physical media will be considered a curiosity.
I dismissed cloud storage , because i don't think a large percentage of people would pay higher bandwidth bills for this feature when they have something like dropbox . And i'm not sure people would pay higher bandwidth bills for internet installs.
I agree, but then in general I don't think that consumers pay more for internet connections for specific features — nor do I think they ever have. I think they pay more because what they have has become a limitation for what they're already doing.
"640K is enough for anyone" (misquote). When there's an excess of computing power of some kind, it may make possible that was previously inconceivable. And if you're the one to conceive and act on it...
An easy one is 3D video (1080p x 2); but I like the idea of omni-video, on a spherical screen with you at the center (or like omniMax, like iMax, but hemispherical so it also covers your peripheral vision - I've experienced it, and the effect is amazing).
Thought expt: what data is inconceivably massive to transmit?
Interesting point, but the thing about eyes is that they can look in unpredicted directions, so you need much more data available than just what the eye can process.
This wouldn't be true if we didn't have lag (then we could just stream the right images for where the eye was looking). But our eyes are extremely sensitive to lag - even done 100% locally, it's difficult to be fast enough (or used to be).
To compensate, we can download a whole environment, and track the eyes locally.
Data that's today inconceivable to transmit usually comes from science. this data is usually being sent via HDD. i don't think it fits the home.
The data from/to the home would be cameras and video streams. i don't know any other sensor technology that generates massive amounts of data and being used in the home.
Maybe some sort of GPU cloud , if the bandwidth is good enough. but usually bandwidth costs more than computing so it would make more sense to gather the GPU units closely.
Omni-video sounds interesting , could you please tell more about the experience ?
I went to the OmniMax theatre at a sort of Technology park in Singapore (about 20 years ago!). I actually haven't seen one anywhere else; the closest is iMax. Anyway, it was a huge hemisphere, enough to cover a large cinema's worth of people. The seats were tilted back by about 45 degrees.
Because my peripheral vision was covered, when the camera tilted, it actually felt like everything was tilting. The movie I saw was taken from a plane, and so this happened whenever it banked. The bandwidth needed was massive - IIRC, they used seven 35mm cameras, and superposed them with seven projectors. The sound was via speakers placed at various locations behind the hemisphere.
I've seen a couple of iMax movies (with an extremely large, though flat screen), and the effect doesn't occur. It's a lot more expensive to make a hemispherical cinema, I think. However, I recall a youtube video of a guy doing it within a simple tent (compensating for the corners not being spherical by distorting the image projected).
BTW: you make a nice point about considering sensors in the home, but also consider what sensors it would be great to have, perhaps using yet-to-be invented technology. eg. a sensor that could register the exact 3D contents of a room; or brain activity; or all your sensory input nerves (brain in a tank anyone?); downloadable objects (auto-milling machines - but probably don't require massive bandwidth, unless it was ridiculously high-resolution, say approaching molecular levels - effectively, matter transmission). An interesting constraint is that lag will probably prevent interesting real-time applications.
There maybe hybrid applications, eg to progressively download massive hires textures in a MMORPG or sandbox game or games like the videodisk-based Dragon's Lair (I think the limit here is that they are so expensive for artists to produce in the first place - perhaps the breakthrough needed is a way for amateurs to generate excellent/interesting textures? Perhaps consumer HD cameras will do this). Maybe motion capture, as in project natal (but I think the bandwidth needed is pretty low... again the problem is latency).
Telstra's subsidiary in New Zealand was also at the bottom of all rankings, but this week they installed the Google cache boxes to mask the issue.
YouTube now hits 3-5Mbps for me. At start of playback it appears to burst a chunk of video, and as it gets close to needing to buffer, it bursts another chunk.
Maybe the Australian parent should look into it, it helps a lot.
I can actually watch 720p and higher without having to pause and come back in 10 minutes.
Gut instinct is that's unlikely to be caused by anything that a cache box can fix as international connectivity isn't as big of an issue on this side of the pond¹.
While I don't agree with the tactic, it is actually a very clever way of reducing bandwidth usage by Telstra. Most youtube videos stream fine at 1Mb/s. I'd guess that a lot of youtube clips are closed before they are over and any buffered data is "wasted".
But Google does a lot of trafic shaping as well: If you Wget a video, you see the first bytes stream in really quickly, after which the speed goes down. No need for your ISP to do this as well.
I assume Google tweaks it so your buffer doesn't get empty, but also not too full: if you decide the close the video, the buffer is wasted (which, on Youtube's scale, is a LOT of data).
That's your only comment?
My voice and accent is what it is. If you want the technical details, it's called a High Rising Terminal.
Show us YOUR video blog and we'll all pass judgment on you.
Dave.
Just wanted to say, thanks for the work you've put into EEVblog. I stumbled upon it while looking for info about the Rigol scope and have been following your stuff ever since. Keep up the good work.