Slashdot Mirror
Netflix Ranks ISP Speeds
Carnth writes "Netflix will start releasing monthly ISP speed reports for the U.S. Google Fiber ranks at the top. They say, 'Broadly, cable shows better than DSL. AT&T U-verse, which is a hybrid fiber-DSL service, shows quite poorly compared to Verizon Fios, which is pure fiber. Charter moved down two positions since October. Verizon mobile has 40% higher performance than AT&T mobile.' Hopefully this will give consumers a better overall picture on how their ISP performs compared to others."
I'm surprised that Google Fiber is large enough to get ranked. I would have guessed that there were other regional ISPs with more customers that weren't listed. Perhaps they're listed simply to encourage the others below them to pick up their speed.
One of my pet peeves as a numerate person not impaired by statisto-phobia is the [ab]use of averages. Sure, the mean contains some information. But the standard deviation contains just as much, if not more! Very seldom do I see anything from which sigma could be inferred, yet whenever you collect data for averages, you can easily calc sigma.
In this case, network averages are useful only for advertising and not much use at all for consumers, with the possible exception of some large corporations who might reasonably suppose they have enough users spread evenly so they _on_average_ will see the average.
For individuals, what matters is the service you will see. And that depends with any carrier more on the neighborhood loading and upstream provisioning on that node.
The only real info you might guess from averages, provided you can make some reasonable assumptions about wirespeed, is what percent of a providers customers are under-provisioned. If cable is commonly 6 Mbps and DSL is 3 and they both net 2, cable is horribly cramped in spite of higher bandwidth.
If they provided a way to narrow it down and see only the ones in your area, it would be very useful. Big ISPs are killing the internet and any sort of consumer guide that presents their information without that of smaller competitors is ultimately a disservice. That said, this information is very useful and interesting, and I would encourage them to continue posting it - just please make it more inclusive. My provider is a small customer-owned co-op and the service is extremely competitive - it would be helpful for that information to be available alongside ratings for the industry giants.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Friends don't let friends enable ecmascript.
Each of us only has one, two, or maybe three (if we're lucky) options to choose from, does it really matter if some ISP that doesn't serve my area is faster than the ones available to me?
My guess is that they just want to ack some pressure on the big ISPs who all want Netflix to cough up for outbound traffic.
FCKGW 09F9 42
and found much the same that the most the remote servers we were using would allow us to pull was about 5MB/s sustained.
Did you rule out the possibility that it was just the end-to-end latency, because the server was far away?
Remember, with the TCP protocol, as end-to-end latency or distance increases: the maximum possible throughput decreases, and the minimum TCP buffer/window size required to achieve the maximum possible speed increases.
E.g. at 100ms round-trip latency, you have to have a TCP buffer size in excess of 256 Kilobytes, to get a throughput of 20 Megabits/Second; which requires special tuning at both ends of the connection.
If your TCP buffers are stuck at 64K; the best possible transfer speed at that latency will be 5 Megabits; even if you have 1 Gigabit of throughput to the server, and the server has 1 Gigabit of throughput to you end-to-end.
Bullshit. The states with higher population densities than European countries still have worse broadband. Americans just accept crap for a high price.
All of Europe isn't a city and we manage to get broadband to rural people. I suspect NJ is more than comparable.