Idea: The 10000 user experiment

guru3 · Post by **guru3** » 2004-03-13 07:29

Do as anyone think that it would be possible to get 10000 users in under 36h on one dc hub?

ivulfusbar · Post by **ivulfusbar** » 2004-03-13 07:58

people is not the problem, the problem is bandwidth which grows has squard growth.

Qbert · Post by **Qbert** » 2004-03-13 08:06

I think a more pertinent question is what is needed on the server and what hub software is best in order to sufficiently supply for 10,000 users.

Otherwise, theoretically I see nothing stopping it from happening.

Qbert · Post by **Qbert** » 2004-03-13 08:20

Qbert wrote:I think a more pertinent question

This was not meant in response to ivulfusbar's comment. My post simply had a connection problem and didn't get posted until many minutes after the previous comment.

Now I have timeouts. I'm glad I can isolate this to sourceforge, because otherwise I love my connection. I hope this doesn't get posted twice.

guru3 · Post by **guru3** » 2004-03-13 15:24

So really ir comes down to bandwith. I better find out exactly how much I have. Ya' think users would possibly put up with a lot of lag to reach the 10,000 mark?

Post by **GargoyleMT** » 2004-03-13 18:00

guru3 wrote:So really ir comes down to bandwith. I better find out exactly how much I have. Ya' think users would possibly put up with a lot of lag to reach the 10,000 mark?

If you over-saturate your connection, users will drop. It's not a matter of "putting up with lag."

Code: Select all

<sandos> does anyone have the rough formula for hubtraffic for X users?
<ivulfusbar> there was (x/20)**2 kbps
<cologic> verliba's.
<ivulfusbar> since 2300-2600 is limit for 10Mbit.

So, for 10,000 users, you do the math.

guru3 · Post by **guru3** » 2004-03-13 21:47

Right, so 10000/20 = 500. 500*2 = 1000. 1000 what? kbps?

Post by **GargoyleMT** » 2004-03-13 22:08

Oops, perhaps I should've said do the correct math. ** is exponentiation.

(10000/20)**2 = 250,000 kbit/s

I found the original chat:

Code: Select all

[2003-08-26 08:14] <Verliba> << Got nice formula for you guys
[2003-08-26 08:16] <Verliba> << resp. U=(N/20)^2
[2003-08-26 08:19] <Verliba> << no
[2003-08-26 08:19] <Verliba> << 220
[2003-08-26 08:19] <Verliba> << that's what formula gives
[2003-08-26 08:19] <Verliba> << I tried myself and 220 was possible
[2003-08-26 08:19] <Verliba> << 1-2s lag
[2003-08-26 08:19] <Verliba> << stable
[2003-08-26 08:20] <Verliba> << I need verifications ;o)
[2003-08-26 08:20] <Verliba> << on czpro formula works too
[2003-08-26 08:21] <Verliba> << (6000/20)^2/1024 ~ 88mbps which is reality
[2003-08-26 08:22] <Verliba> << the upload is  in average 64 but lot's of time it goes to 92
[2003-08-26 08:26] <Verliba> << i put this to docs on sf.net
[2003-08-26 08:27] <Verliba> << later i'll try to make a table of users/cpu
[2003-08-26 08:27] <Verliba> << that isn't so easy to estimate

Verliba is the author of Verlihub, which seems to have pioneer the recent large-user trend.

guru3 · Post by **guru3** » 2004-03-14 06:50

Got it, but in standard math ^ is for exponents... (I see ^ and I think "to the power of...")

cologic · Post by **cologic** » 2004-03-14 13:49

You're wrong.

Even within computer math, virtually none of the C-based languages use ^ to signify exponentiation (they don't tend to have an operator for it). Mathematica appears to use it, as does BASIC. Fortran uses **.

It's "in standard math" not the exponentiation operator, either; I recall from no mathematical book I've read such a convention. Rather, they tend to use superscripts.

So, yeah, you're wrong.

Qbert · Post by **Qbert** » 2004-03-14 14:57

cologic wrote:You're wrong.

I don't know. When I see published mathematics textbooks, especially when used in schools or universities, I think very reputable of them. And when they sometimes reference calculators, especially TI calculators, I've known them to give "^" as information on how to perform exponentiation. They don't only say it to explain how to use your calculator, but also for sample TI-BASIC programs they supply in the text or when explaining how you can use a CAS (not just TI) to solve a problem.

But besides all of that, I regard TI highly and reputable because I love to use my TI-89 and programming (assembly and flash) for it. When I see "Exponentiation ^" in their guidebooks, I put it in my mind as being official.

Qbert · Post by **Qbert** » 2004-03-14 15:04

cologic wrote:Even within computer math, virtually none of the C-based languages use ^ to signify exponentiation

Just because a programming language (specifically a C basic language) doesn't use a specific operator for mathematics, that doesn't mean it's not standard math. How about factorial !, which is standard math.

cologic · Post by **cologic** » 2004-03-14 15:56

And when they sometimes reference calculators, especially TI calculators, I've known them to give "^" as information on how to perform exponentiation. They don't only say it to explain how to use your calculator, but also for sample TI-BASIC programs they supply in the text

So .. when it refers to programming languages, which you already discounted as authoritative sources of mathematical notation. Right then.

How about a published mathematics textbook when it's referring to the math, and not trying to translate it to calculators' notations?

guru3 · Post by **guru3** » 2004-03-14 16:53

I don't really care what the correct mathmatical notation, it's beside the point. I didn't understand what you meant, it has now been clarified, let it be.

Qbert · Post by **Qbert** » 2004-03-23 08:09

I just found this out in my Mathematica class, and was reminded of this thread. Just as you said ^ is used in Mathematica as power, it still also has a **. This could just add to the confusion that different areas use the same accepted symbol for different tasks.

NonCommutativeMultiply (**)
a ** b ** c is a general associative, but non‐commutative, form of multiplication.

DC++

Idea: The 10000 user experiment

Idea: The 10000 user experiment

Who is online