GeoffM

That entirely depends on his connection... and the one on the other end. He implied that his connection wasn't good.

10ms/80Mb/s here and 157ms/45Mb/s to Newbury for me. Both are perfectly adequate for streaming multiple video feeds at once.

Not all video uses TCP - most uses UDP, though apparently Netflix does use TCP. I would guess that's to do with being more "local" than most Internet streams. UDP is better for video generally because it doesn't care about lost packets - much of the time you wouldn't even notice a single, or even a few lost packets. For the same reason, packets arriving later in sequence can silently be ignored. To achieve a moderately reliable throughput with TCP requires a lot of buffering - basically "catch up" time for recovering lost packets. UDP is fairly instantaneous. I also note Slingbox uses TCP but can also use UDP, but bear in mind it's not just a pure video feed but control data too.

On a side note a few years ago I was listening to a short article about BBC radio in the UK where the broadcasts are made from one location but transmitted (obviously) in various parts of the country. Some of the channels have time signals so you can set your watch by them. Since the broadcast station to transmission towers apparently use digital - with the resulting delays - the transmission in London (or Manchester, or wherever it is) is actually broadcast a few seconds early so that the time signal can be transmitted locally at exactly the right time. Well, I thought it was interesting anyway!

Michael

I agree if it remains at that speed but the server that the OP has is a home connection and it may possibly have an unbalanced connection on the server side.
I agree that UDP is fast but if it is used for video services, all video decompressors rely on the previous screen to determine how to produce the next screen. Therefore if a single packet is lost, the screen can start to break up (especially action sequences) and can get significantly worse until a sync packet is sent and for action sequences, that can take several seconds.

When you refer to video, are you referring to video or video services? If you are referring to video services, that would really surprise me because of the experience that I had using UDP since backbones lose packets all the time since if more data is pushed into a backbone than can be handled, packets are thrown away. However even YouTube seems to be using TCP/IP since I don't see missed packets or breakup but only screen slowdown when on a very slow connection and the same with porn sites.

I've programmed with UDP and I found it not to be reliable. I had a workstation program where I wanted to determine the status of a bunch of servers across the internet at a regular interval. I used UDP and sent a packet to each of the servers and sometimes I got a packet back from each of the servers and other times I didn't when all the servers were up and running the application that I was concerned about. Finally I sent out about 10 UDP packets in a row to each server and as soon as I received a packet back from a server, I updated the display status for that server and ignored the remaining UDP packets that that server may send back.

GeoffM

Yes, it needs regular full frames (I frame?) but the beauty of most video is that you can get away with losing a bit here and there until the next full frame comes along. The exception being, as you note, action or complete changes of scenery.

TBH I don't know what you mean wrt video or video services. Most of my programming is desktop with LANs (think simulators and CCTV) so the big wide world is somewhat more extrapolated than our lab stuff.

Push too much data onto a TCP stack and it'll lose packets too... despite what people say. After all, there's only so much buffering it can do before it's full and packets time out. It's a fun argument to have with the uninitiated. (The other one is when they fail to realise that TCP is a stream... not a "pump 2000 bytes in, get 2000 bytes out in one Read()" kind of system)

As for porn, well, it is said that the adult entertainment industry was a key player in video streaming...

Yes, most of my work is TCP because lossy UDP is not helpful. I have used it, but only for stuff that either updates everything frequently, or it doesn't matter if stuff gets lost.

Michael

When I refer to video services, I'm referring to web sites like Netflix. Amazon Prime, BBC, PBS, etc. which play long playing videos. Most of the other stuff, I really don't care if the video gets messed up occasionally.

However when I first moved into my condo, everyone had "g" type routers and about 5 pm, I'd start watching a Netflix movie and about a half hour later, Netflix would pause and start buffering. It took me a long time to figure out what was happening but then I realized that Netflix checks my connection speed to determine the quality of the video to stream but my WI-FI speed was slowing as people got off work and started using their systems in the complex. So I purchased a single band "N" type router and still had the same problem. Finally after scratching my head, I realized that it didn't make any difference that I had a high speed connection on the 2.4 GHz band since the air time was being used by other people's "g" type routers.

I returned the router and bought a dual band router and put all my connections that supported the 5 GHz band on that band (including my built in HDTV streamer) and the problem was solved.

Michael

I agree but TCP/IP does retires. If a video service tests the speed of a connection, fills a look ahead buffer before it starts to play a video, and streams video quality that has a margin of error for the connection (connection speed is 15 mbps but 10 mbps quality is streamed), retries normally shouldn't have an effect unless there are so many that the look ahead buffer empties.

GeoffM

To a limited extent. What I'm saying is trying to pump even theoretical near maximum bandwidth through a pipe is not going to end happily because over time it will simply run out of room and time.

GeoffM

Are you saying it was something other than too many devices trying to communicate over the same channels? I've heard of this problem before in condos where wifi routers are too near each other and are constantly colliding and backing off. Even with channel hopping, only a finite number of channels. I suppose that's one benefit of living out here - right now I'm seeing just 2-4 other routers on wifi, with 2 of those dropping out frequently so unlikely to be much bother.

Michael

It was just massive WI-FI collusions that caused my throughput go from about 16 mbps to about 2 mbps.

"N" and "b,g" protocols only have 3 non-overlapping channels (channels 1, 6, and 11) for the US 2.4 GHz band if the router has the "N" protocol and uses 20 MHz wide channels. The "b,g" type routers only use 20 MHz wide channels. To be 802.11n compliant, the router must default with 20 MHz wide channels on the 2.4 GHz band to allow people to have 3 non-overlapping channels on the 2.4 GHz band but that means that if they have an N300 router and N300 network adapters, they are not running at N300 link speed but 130 mbps maximum link speed which is what the table in the 2nd picture is indicating. I wasn't sure I believed the table until I looked at the networks that I see in my condo unit and sure as hell, there were "MAX DATA RATES" that I didn't expect on over 95% of the networks that I see in my unit. The third picture is a snapshot of sorted networks by the slowest "MAX DATA RATES". In the picture you can see the that there are 3 "b,g" type routers with a "MAX DATA RATE" of 54 mbps, 1 "b,g,n" router with a "MAX DATA RATE" of 65 mbps, 1 "b,g,n" router with a "MAX DATA RATE" of 130 mbps, 8 "b,g,n" routers with a "MAX Data RATE " of 144 mbps, and 11 "b,g,n" routers with a "MAX DATA RATE" of 217 mbps. There are also several more pages of routers that I can see from my unit.

54 mbps makes sense for the "b,g" type routers, and the 65 mbps, 130 mbps, and 217 mbps are all in the first table indicating that router to network link speed is running slow because the channel width is only 20 MHz. The 144 mbps link speed is from the second table when using 2 spatial streams and a 20 MHz wide channel. I believe the difference between the tables is top table has two antennas and the both has two spatial streams for N150 and N300 devices but both are suing using 20 MHz wide channels. By looking at the lower left, you can see that routers are sitting on the top of each other and if you notice, there is a network name of TigerTime2GHz that is using a wide channel. That's my channel since I set my router to use a 40 MHz wide channel on the 2.3 GHz band and my router indicates a "MAX DATA RATE" of 300 mbps but if I reduce the channel width to 20 MHZ, the "MAX DATA RATE" drops to 130 mbps. So my 40 MHz wide channel is interfering with just about everyone but I seldom use the 2.4 GHz band and I use my 5 GHz network (TigerTime5GHz) to use a 40 MHz wide channel starting at channel 44 and at that channel, I'm all alone.

You'll notice at the lower right that only one router is using channels 48-140 on the 5 GHz band and that is because many routers don't support those channels since they are DFS channels (radar channels). DFS channels are not hard to support but they take forever to come up and the router must constantly monitor for radar signals and move to another channel if radar is detected and if it moves to another DFS channel, it takes forever to reestablish the network connection since the router has to monitor the channel for radar signals for a long time before using it.

The final picture is a table of maximum link speeds on the 5 GHz band when using 40 MHz wide channels for N devices. The 80 MHz wide channels for the "AC" protocol are unimportant when referring to the 2.4 GHz band but the maximum link speeds for N150 and N300 devices will be what is indicated if 40 MHz wide channels are used on the 2.4 GHz band.

Channel hopping is not used by WI-FI on the 2.4 GHz band. Channel hopping is used by Bluetooth on the 2.4 GHz band.

Macleodoftheclan

I live in Long beach California. There's only one Internet provider in this area! Does any one else find that as ridiculous as I do? Any way it's about $70 a month. Which just like mobile phones over here is way over priced compared to the UK.

Michael

Link speed is not necessarily the criteria for throughput. For N300 routers, you can see that the maximum downlink speed varies significantly depending on the router from about 25 mbps to 89 mbps depending on the manufacturer and model number of the router. The second chart gives typical maximum speeds of "N" and "AC" router on the 2.4 GHz and 5 GHz bands. None of the routers come close to their advertised link speed which is in the third chart.

Michael

No considering that the lower 48 states has over 30x the area but services only a little more than 4x the population plus the US cable companies laid high capacity cable systems that's typically 2x or possibly more the capacity of UK cable TV systems. Because of that over capacity (expensive infrastructure cost), both Verizon and AT&T saw an opportunity to make money against the cable TV companies and each started their cable systems but Verizon abandoned it's plans after very limited cable laying since they couldn't get enough customers to make money with fiber directly to the homes. AT&T which already had fiber in the streets in many areas went with a lower cost system which used twisted pair for the last mile but that limited internet speeds significantly. Some think Google cable will be the solution since Google is laying 1 gigabit fiber in the streets instead of 4-7 gigabits laid for cable TV systems but Google has to pay for that infrastructure with only internet service and once they lay cable in expensive areas, that may not be possible. The difference in cost between a 7 gigabit fiber system and a 1 gigabit fiber system is much more expensive fiber cable and 7x as many amplifiers in the streets.

If in the future if people no longer want cable TV (about 2/3rds of the US TV market but declining), cable systems will go bankrupt unless they can figure out how to use that excess capacity and then something like Google cable with it's low infrastructure cost will be viable. Not only do US cable TV systems currently have plenty of capacity since analog was dropped, but if QAM 4096 is used for the entire cable system (has already been specified for DOCIS 3.1 cable modems), that would increase the capacity by 40%. Cable systems could also free up more bandwidth by using better compression/decompression (MPEG-4 instead of MPEG-2) for TV. If cable companies go bankrupt, a fire sale may have to be below the cost of laying a Google like cable system unless someone can figure out what to use the capacity for (possibly dedicated OC connections for large businesses in cities) since the cost of maintaining all those amplifiers in the street would be an additional cost.

As far as mobile phones, someone has to pay for all the cell towers is the US. If cell providers were making an unreasonable profit, others would jump into the market but instead Deutsche Telekom has been trying to sell T-Mobile for years since it hasn't been profitable. AT&T tried to buy T-Mobile to expand it's network but was blocked by regulators and Sprint made an offer for T-Mobile but the offer was withdrawn when the cost didn't make financial sense. Verizon, AT&T, T-Mobile, and Sprint could sell off their cell towers to a consortium and each lease the towers which in theory should reduce costs but regulators wouldn't allow that since then the cell tower owners would have a monopoly and would have to be regulated and in the past, any private regulated monopoly increased costs.