Read on Twitter
I have Opinions about the #discoverbc website mess. As someone who did site performanc testing for, like, the Olympics that one time, the most difficult problem isn't technical, it's psychological: your tests have to be based on realistic usage scenarios. This is where most fail.
My guess is they didn't expect the flood right at 7:00am. /\\ is a hard profile to provision and plan for, since it puts you in a while new league even if only for a short while. Throwing hardware at it isn't often helpful on the fly if there are integration bottlenecks.
I dealt with a sports festival where we had about 70k visits in the first 10m, with an ecommerce, ticket and mail system in the mix; we had to do multiple rounds of testing to clear each bottleneck, nevermind just sizing up hardware. It's a process.
Performance aside, looking at the homepage of the DiscoverBC site tells me they didn't really think through how people would interact with the site in other ways also - the UX needed way, way more testing too.
But the key thing is to think about how likely people are to complete each step, based on last year's traffic and interaction studies. Then simulate those at estimated max load+a bit, then add error and background traffic. Then see how things survive, fix, then do it again.
Just flooding a server with traffic does not work. The results become meaningless noisy failures. But if you get the estimates or scenarios wrong... you are testing the wrong things, and you won't see the failures until it's too late. Like this morning.
