USA ISP Speed Index • Hacker News Board

Re: USA ISP Speed Index

#21

Post by brixon » Tue, Jan 14, 2014, 10:35 PM UTC

ancarda wrote: post #20

I don't understand how this is measured. For instance Google Fiber is 3.69 Mbps. For a gigabit internet service, it seems fairly low.

The top number might be the most that Netflix needs to stream a video. The numbers are not important, this is more about variance to each other.

Re: USA ISP Speed Index

#22

Post by penguindev » Tue, Jan 14, 2014, 10:40 PM UTC

modeless wrote: post #19

penguindev wrote: post #2

hmm. comcast dropped from 2.17 to 1.63 mbps YOY. awesome. luckily, speedtest.net shows me a lot higher than that... 25mbps, to be precise. Edit: I see Netflix has tweaked their results to use 'Prime Time' readings only. I guess I should retest this evening as well.

I care about my bandwidth to Netflix a lot more than I care about my bandwidth to speedtest.net.

You can test on many different servers, I tested to a server in virginia since I assume it would be a reasonable proxy to ec2 us-east bandwidth.

Re: USA ISP Speed Index

#23

Post by Rebelgecko » Tue, Jan 14, 2014, 10:44 PM UTC

ancarda wrote: post #20

I don't understand how this is measured. For instance Google Fiber is 3.69 Mbps. For a gigabit internet service, it seems fairly low.

It looks like this is just measuring the average bitrate that people are streaming at. It won't reflect whether or not someone is capable of watching higher quality streams than Netflix is actually giving them. I don't think their highest quality is more than 6 megabits/sec, and it's only available on certain devices and movies.

Re: USA ISP Speed Index

#24

Post by geraldcombs » Tue, Jan 14, 2014, 10:46 PM UTC

RKearney wrote: post #11

geraldcombs wrote: post #9

Why no latency measurements?

My guess is because latency is not relevant when it comes to streaming videos from Netflix. A consumer will not be able to notice their video took 40 milliseconds longer to start streaming. They will, however, notice when the video stops to buffer every 10-15 minutes because their ISP can't keep up with the throughput.

Would latency affect the non-streaming parts of Netflix, e.g. browsing titles?

Re: USA ISP Speed Index

#25

Post by X-Istence » Tue, Jan 14, 2014, 10:46 PM UTC

penguindev wrote: post #22

modeless wrote: post #19

Earlier quoted context omitted.

I care about my bandwidth to Netflix a lot more than I care about my bandwidth to speedtest.net.

You can test on many different servers, I tested to a server in virginia since I assume it would be a reasonable proxy to ec2 us-east bandwidth.

Netflix hosts their video servers much closer than ec2's us-east...

Re: USA ISP Speed Index

#26

Post by ghshephard » Tue, Jan 14, 2014, 10:50 PM UTC

thisiswrong wrote: post #4

penguindev wrote: post #2

hmm. comcast dropped from 2.17 to 1.63 mbps YOY. awesome. luckily, speedtest.net shows me a lot higher than that... 25mbps, to be precise. Edit: I see Netflix has tweaked their results to use 'Prime Time' readings only. I guess I should retest this evening as well.

False. Speedtest.net gives results in Mbit/s. 25 Mbit/s ≠ 25 Mb/s. 1 Mbit = 0.125 MB

While others have correctly pointed out that 25 Mbits/s does in fact equal 25 Mb/s, (though, I always use Mbits/s, just to eliminate the confusion) - I'd also like to note that bandwidth is always measured in SI Mbits - I.E. 25 Mbits/s of bandwidth is always 25,000,000 bits/second.

Any time you see someone call it out explicitly as 25 Mib/s, double check and see if that's what they actually mean - 90+% of the time it isn't.

Re: USA ISP Speed Index

#27

Post by ghshephard » Tue, Jan 14, 2014, 10:51 PM UTC

avis wrote: post #15

this seems more of a breakdown of the avg customer link speed/ isp than an isp performance chart.

It's actually more a breakdown of Netflix -> ISP -> Customer performance chart.

Re: USA ISP Speed Index

#28

Post by modeless » Tue, Jan 14, 2014, 11:01 PM UTC

penguindev wrote: post #22

modeless wrote: post #19

Earlier quoted context omitted.

I care about my bandwidth to Netflix a lot more than I care about my bandwidth to speedtest.net.

You can test on many different servers, I tested to a server in virginia since I assume it would be a reasonable proxy to ec2 us-east bandwidth.

The list of speedtest.net servers is public and I don't doubt that Comcast uses it to drive traffic shaping decisions. Speed tests are meaningless except to establish an upper bound on last mile speed. They are not representative of bandwidth to real services like Netflix, YouTube, Dropbox, Steam, etc.

Re: USA ISP Speed Index

#29

Post by kirkbackus » Tue, Jan 14, 2014, 11:03 PM UTC

ancarda wrote: post #20

I don't understand how this is measured. For instance Google Fiber is 3.69 Mbps. For a gigabit internet service, it seems fairly low.

It is a gigabit internet service, but there are also people on the free internet plan which is slower.

Re: USA ISP Speed Index

#30

Post by kgp7 » Tue, Jan 14, 2014, 11:06 PM UTC

penguindev wrote: post #22

modeless wrote: post #19

Earlier quoted context omitted.

I care about my bandwidth to Netflix a lot more than I care about my bandwidth to speedtest.net.

You can test on many different servers, I tested to a server in virginia since I assume it would be a reasonable proxy to ec2 us-east bandwidth.

I am pretty sure nothing is streamed off of AWS - possibly Cloudfrount , but based on what I know of Netflix use L3 and Akamai heavily.