This is a good point, but it could be somewhat mitigated by their expected growth of up to 2M per second. To quote the full line from their blog post:
> Currently our production load peaks at around 700K events per second. To account for the future growth and possible disaster recovery scenarios, we settled on a test load of 2M events per second.
In the absence of any other numbers, using the 700k number seems reasonable otherwise you would start to get into modelling without any real data to back up your assumptions. This at least gives you a maximum they would be paying under current production conditions.
That's actually a reasonable response. From my first read the numbers seemed a bit inflated, so I dug into their blog post out of curiosity (I develop on GAE), so the comment was not ill intended. Now that you have pointed out they do seem a good estimate!
Thanks for the great articles David, keep them coming! :)
If this is the logic, then a better title would have been "How much will Spotify be paying Google Cloud in a few years, assuming 3x growth from today's numbers?"
It's really not fair to tell the world that their costs are based on average of 700k rps, when that is their peak traffic through the system.
You could guess that, but with no real expectation that it's right either. E.g. for a global service I'd guess it's better viewed as a sum of three offset distributions, one per continent, but even then the individual distributions are unlikely to be gaussian.
Having run global services myself I can tell you that the shape is usually no standard distribution. Peak/trough depends on the exact usage demographics, but the one I have on hand has a trough of ~1/2 peak, and an average of ~3/4.
> Currently our production load peaks at around 700K events per second. To account for the future growth and possible disaster recovery scenarios, we settled on a test load of 2M events per second.
In the absence of any other numbers, using the 700k number seems reasonable otherwise you would start to get into modelling without any real data to back up your assumptions. This at least gives you a maximum they would be paying under current production conditions.