correlation of speedtest.net data and actual speeds

[MrDarceh]

I've been having net troubles of late so, whilst in the process of playing around with my router, I thought I'd do a speed test. I'm allegedly running at around 1.5mb/s, but my connection generally runs at around 170kb/s. Is this at all unusual?

[Primüs]

there are a lot of terms and abbreviations to be weary of.

There is mb/s which would indicated MegaBITS per second, and mB/s which would be MegaBYTES per second.

Same with kb/s and kB/s.

However, people tend to write things wrong.

Speedtest.net tend to show in mb/s or megabits per second.

Considering there are 8 bits to a byte, also taking in to account overheads and the minimal variance of converging kilobits to kilobytes, you could say 1.5mb/s is 1500kb/s, which would be around 150kB/s.

So around 1.5megabits / second, would sound about right to average 170 kilobytes per second.

[mshagg]

Do you mean 1.5Mbit/s and 170kB/s? If so then that is about right?

[MrDarceh]

derp. thanks guys

[tin]

Originally Posted by Primüs

However, people tend to write things wrong

Yeah - The mainstream media are pretty bad for it. Like recent headlines about a 1 gigabyte per second network being rolled out in Kansas City... Almost every article I read about it kept saying gigabyte (including "quotes" from Google reps).

[caspian]

Originally Posted by MrDarceh

I've been having net troubles of late so, whilst in the process of playing around with my router, I thought I'd do a speed test. I'm allegedly running at around 1.5mb/s, but my connection generally runs at around 170kb/s. Is this at all unusual?

the other annoyance about speed tests is that they test *throughput* from a remote data source, but then claim it's your connection speed, which too many people then think they can correlate with their DSL sync speed.

in reality, DSL sync speed merely sets up upper limit for the throughput that can be achieved, and only then after every overhead has been taken into account - throughput can be significantly lower.

[MR CHILLED]

How accurate is speedtest anyways?

[evilasdeath]

Originally Posted by MR CHILLED

How accurate is speedtest anyways?

how long is a piece of string.

Speedtest isn't a true representation of anything since you don't know how it's coded in the back end to handle the transfer and how it calculates it.

Could be done in UDP or TCP or just a stream of data inside another session with a modified TCP stack.

People really need to understand where problems exist in the internet.

The access tail is only 1 part sure most problems are there but what many don't understand is everything past the first access-router in the ISP is now the Internet and its entirely best effort, you don't get special treatment over any other traffic. There are so many factors in play between any 2 hosts on the internet. And many of them rely on routing by dollar signs $$$.

[caspian]

Originally Posted by MR CHILLED

How accurate is speedtest anyways?

put it this way: put 1000 random drivers in their cars and have them drive from your house over a random route to a random McDonalds at a random time of day. when they get there, they order a cheeseburger and take another random path home.

when they get home, divide the distance travelled by the time taken to make the journey.

do you think the result means anything, or is it just random rubbish?

the drivers have no standardised capability of driving.
their cars will have different levels of performance.
they drove different distances.
they took nonstandard routes - were they suburbs full of traffic control or freeways? or in the middle of nowhere? even different drivers driving to the exact same destination may have taken very different routes.
they drove at random times - was the journey in peak hour congestion or at 3am?

the cheeseburger is the only standardised part of the test, and even then there's no allowance for how busy the store was when they got there.

that's about as much as speedtest or any web based test means.

at best, it demonstrates what your connection acheived under random circumstances - but you measured the time taken sitting in your lounge room with zero appreciation of any of the variables in test, and generally while complaining you're hungry and why won't they hurry up.

at worst, speedtest then purports to use the result generated to tell you how fast you can back down your driveway.

btw, your cheeseburger is cold, and that youtube clip is still buffering too.

[MR CHILLED]

lol.....ok ok....we get the message...heh

I want a cheeseburger now : (

[caspian]

I'll bring you one over. I may be a while, but you don't know I'm coming from Melbourne - even if I drive like the clappers you won't be able to distinguish the result from Granny McGoo coming from across town. the speed with which the burger arrives will be the same.

luckily, I have a low latency cheeseburger CDN fairly close by, might be time for a multithreaded test myself later.

[edit] I just need to talk my download manager into going for a drive for me, I've had a few brews and corrupted my TCP stack a bit too much.

[Pugs]

Originally Posted by MR CHILLED

lol.....ok ok....we get the message...heh

I want a cheeseburger now : (

ditto.. you're bad man mister caspian

[pirateRob]

Originally Posted by MR CHILLED

How accurate is speedtest anyways?

speedtest is useful. Mostly in the sense that if I go onsite to test a new 10mb/10mb connection and I get 90% or more of that speed, I can say yeah it's working as I expect. However if I get 5mb after numerous tests, then time to complain to the ISP.

Home dsl is a bit harder to determine as there are more factors that affect overall line quality, in those cases it gives you a good guide.

"Guide" is the key here. It's not perfect nor can it be. Those series of tubes are complicated.