Also, experiments at this size and speed have never been carried out before.
And statistical behaviours are very difficult to understand. First thing: 99.9999% uptime for ALL users is HUGELY different from 99.999%.
As a matter of fact, this was just one of amazon’s zones, rememeber.
Edit: finally, the right model for these systems might well have no mean (fat tails…) and then where do the statistics go from there?
Also, experiments at this size and speed have never been carried out before.
And statistical behaviours are very difficult to understand. First thing: 99.9999% uptime for ALL users is HUGELY different from 99.999%.
As a matter of fact, this was just one of amazon’s zones, rememeber.
Edit: finally, the right model for these systems might well have no mean (fat tails…) and then where do the statistics go from there?