US Average upload is over 8Mbps?

[mudmanc4]

Something is skewing the results? Likely the several recent wide connections?

 

The question is, then to keep the results sane how to detect a higher level throughput and in some way separate this from the average. 

 

 

So which one is it, 8.4 upload or 32.4 ?

[CA3LE]

That's because some guy was blowing 100 Mbps+ upload results recently (in relation to your query). Look on the stats page on the right under the graph. It's considering the last 50 of each type. You can change that. The results page is averaging more than the last 50, currently set at 250 (haven't changed that in years but it may in the future).

The results page is lagged a little from the actual stats pages (which are live). The results page can be switched to live mode and I'll definitely make that an option in the next versions (which will have an options control panel). A long while back I made a program that runs through every couple minutes and pre-runs all the numbers so when people load the results page it doesn't actually have to do the calculations every time for everyone. Made results load much faster... but to be honest on this server I can switch it to live mode and barely tell a difference. Where the difference matters is on scale. If it were the old way when the massive amount of reddit traffic hit I think the server would have overloaded under half the traffic it handled. Those calculations add up. But thinking about it, I can make it an option people could toggle. As long as the wide scale is on the more optimized version it's all good. I just looked at the code and it's just a simple variable to switch it to the old live version. ...why not let people access that? :-/

So to answer your question... it's both. Both are right, the question asked is what changed. --- but also realize because the results page stats run a couple minutes behind so even if you set the stats page to 250 results it won't match. You'd have to refresh the page at the exact time the compiler program comes around to run calculation on the corresponding database. Pretty much impossible timing. Less popular stats pages will match because less changes over the period of time.

[mudmanc4]

I hear what you are saying, so considering were getting to 'that point', where people are getting split wide open connections for those results at and over 100Mbps Tx, anticipating this will take quite some time before even a noticeable percentage of connections can run like this, implementation of a detection device might be presentable. 

 

Say a connection comes in at +/- 50% + to the average, program places this in a special database, which should have built in ability to be merged in say X# of years when many more are testing in that range. 

 

This way the results stay clear and real. Does that spark anything off in you? 

[CA3LE]

Wouldn't really slow results offset the really fast ones?  In theory yes, in reality... no.

 

Example

(5+5+5+5+5+5+6+6+6+6+6+6+150+0.01)/12= 18.00 ... vs 6.6 if the high and low were removed.  You're right.  They should still be included and calculated but what I could do instead is make a new calculation we could call the middle average (or something).  The new average will take the heaviest outlying numbers out of the equation.  I'll give people the ability to toggle this as a setting when I open the preferences panel to the public.  ... the new algorithm will most likely end up being the default.