This point is not so simple: > (and unfortunately may remain > that way because instance optimization is markedly revenue and profit > negative for GAE)
It really is not in google's economic interest to have idle instances running. More idle instances means more infrastructure cost, and, if you charge for it, more user backlash. Idle instances are a necessary evil when you want responsive, dynamic scaling. But technically, you only need as many as it takes to satisfy the simple equation (rate of join * max wait) / startup time. That is, you need to have enough on hand to handle the incoming requests while you spin up some more. The key problem is that for probably all applications, the numerator in that equation is completely different for tasks than it is for users browsing the site. And it is often different for different kinds of tasks. (In fact, it's often different for different users/urls/handlers as well, but sometimes too much configurability is a bad thing.) The scheduler isn't as complicated as many people seem to believe. The only real magic in it is the heuristics for computing the three values in that equation when the user sets them to "automatic." The rest is simple trend averaging. -Joshua -- You received this message because you are subscribed to the Google Groups "Google App Engine" group. To post to this group, send email to google-appengine@googlegroups.com. To unsubscribe from this group, send email to google-appengine+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.
