Another Speed Testing Site

[ColinTaylor]

Seems to me isps like to help tcp as much as possible but are causing grief for udp.

That's probably because UDP is an unreliable protocol by design.

https://en.wikipedia.org/wiki/User_Datagram_Protocol#Comparison_of_UDP_and_TCP

 

[el pescador]

Ive read up way back...thats how t feels.
how do i measure?
whats causing my udp to go bad?

Is there anythign other than iperf around for a simple udp test on windows?

 

[Nullity]

That's probably because UDP is an unreliable protocol by design.

https://en.wikipedia.org/wiki/User_Datagram_Protocol#Comparison_of_UDP_and_TCP

Being unreliable has little to do with throughput though, right?

I mean, if we are measuring IP bitrates, whether it's TCP or UDP is moot.

Perhaps if we are measuring "goodput", the protocol overheads become important, but is that what we are trying to measure? I guess it depends on the observer's objective.


Personally, I've rarely ever had speedtests yield inaccurate results but I've only had a few ISPs throughout my life.

 

[el pescador]

All speedtesting sites dont use udp.
So your only testing a few parts of our internet.

I think since net neautrality they got around throttling p2p etc by simply throttling one protocol .

Correct me if im wrong....but thats how it feels.
And they wont be kicking tcp now will they....it will simply resend.

Also the website that checked whether your isp was throttling was taken down a few months ago so its not easy to check.

Whats going on there?Brown envelope job?

 

[ColinTaylor]

Being unreliable has little to do with throughput though, right?

I wasn't really referring to throughput as such, rather his statement about "isps like to help tcp as much as possible".

TCP guarantees delivery of the data because of it's bidirectional handshaking, and can cope with congestion. Whereas UDP doesn't guarantee delivery and has no congestion mechanism (it has to be done by the application). I like to think that with UDP the sender is just "throwing" packets at the receiver with no real sense of whether they're getting there.

With a TCP speedtest it's easy to measure throughput because you just send a large file and see how long it takes. With a UDP speedtest you have to start off sending packets slowly and gradually increase the speed until you start seeing packet loss. At that point you have to start making assumptions about what is an "acceptable" level of packet loss.

So I don't think it's fair to say that ISP's are "helping TCP", it's just the nature of TCP vs. UDP.