View Single Post
Lt. Commander
Join Date: Dec 2007
Posts: 120
Please consider this graph: http://i49.tinypic.com/scg6zo.jpg

This is an estimation I've made of the server load depending on the hour of the day.
Hours are GMT, you need to convert them to your timezone.
The lower the curve, the lower the server population, the smoother your play will be. If this is estimation is accurate in any way of course...

Disclaimer: It's totally hypothetical and the data is not official or even real, it's all out of my appreciations of some things based on experience and a bit of common sense (hopefully). Please feel free to judge the result, it's meant to be a discussion and I don't come here bearing any truth. just ideas and hypothesis.
This is not science!



So here are the parameters I used.

I considered 4 different periods of the day. Each of them is assigned a value from 1 to 4, depending on how many people are averagely connected at that time.
It looks like this: http://i48.tinypic.com/2uyrjw6.jpg
  • 1 = 01:00 to 10:00 (low population)
    the low pop. corresponds to night time and early morning.
  • 2 = 10:00 to 12:00 / 14:00 to 17:00 / 00:00 to 01:00 (medium population)
    the medium pop. is mornings, afternoons, early night
  • 3 = 12:00 to 14:00 / 17:00 to 19:00 / 22:00 to 0:00 (high population)
    the high pop. is lunch time, late afternoon/early evening and late evening
  • 4 = 19:00 to 22:00 (maximum population).
    the maximum pop is what we refer to as "prime time", a.k.a when most people are free
The real ratio between the number of players connected during low pop. and max pop. is likely to be more than 4. But the point here is to assess things relatively and not as precise quantities. However it would be better to have more realistic coefficients here. Short of knowing how to do that, I assigned simple entire values. But honestly, it kind of messes up the rest of the study, so from this point on it can't be precise. it's just to get a rough picture of things, really.

Then I shifted that classic cliche schedule back or forward in time over 3 major population areas for sto, in order to see what was happening in all three areas simultaneously, at each hour of the day: (It looks like that: http://i48.tinypic.com/1418g3k.jpg)
  • NA, from GMT -8 to GMT -5 (total pop 350M)
  • EU, from GMT 0 to GMT +1 (total pop 300M)
  • Aus, at GMT +10 (total pop 20M)
Then in order to add them hour by hour I needed to balance the different areas. The major unknown is "how many % of each total population plays STO?". Since ST is fairly more popular in USA than it is generally in Europe, I tend to think it's a 3/2 ratio in favor of NA. Australia is small, I added it in to give more realism to the graphs, considering NA and EU are close timezones and there must still be players from all over the world.
I used a fixed ratio of 1:15 between Aus and NA (this is the real population ratio between them), assuming these are fairly close cultures susceptible to buy/play STO equally. Hence it's just the EU balancing that is varying on those curves.

I used three balance settings, adding coefficients to each zone "value" (1 to 4) for each hour of the day.
  • BIG NA: [NA=15] [EU=10] [Aus=1]
  • NA=EU: [NA=15] [EU=15] [Aus=1]
  • BIG EU: [NA=15] [EU=20] [Aus=1]
Plus a fourth reference "all equal" settings with all areas at 1, to see how coefficients affect the curve.

This is a lot of estimation, the data is not rigorous, so one needs to observe things in a fuzzy way.

Nevertheless I think it could be somewhat accurate enough to reflect how you feel things from your unique client blind side. Hours are given in GMT, and all curves have been indexed to 1 for their lower value.

Again, feel free to dismiss some of my ideas/methodology and propose your ideas to improve it theoretically.
Practically, I'd appreciate some feedback, if this thing works a bit from your perspective of server load, or if it's complete nonsense!