The performance monitor is useful for identifying the server load, and being able to request or release extra nodes that it requests to handle burst.
While it may be difficult when to predict a burst will happen, I assume that the monitors assume that if they see a rapid increase in demand that an increased demand will persist so they should request more servers. Otherwise the only way to know that a more intense load will come is if you have some outside knowledge about when people will visit a website.
Not that it matters this far past Assignment 4, but one possible implementation of the performance monitoring my partner and I were wondering about was taking a derivative to see the change in number of requests over time. It didn't seem to work (or we didn't do it right), but I'm still convinced that this is a viable implementation. Are any performance monitors implemented that way?
The performance monitor is useful for identifying the server load, and being able to request or release extra nodes that it requests to handle burst.
While it may be difficult when to predict a burst will happen, I assume that the monitors assume that if they see a rapid increase in demand that an increased demand will persist so they should request more servers. Otherwise the only way to know that a more intense load will come is if you have some outside knowledge about when people will visit a website.
Of note: https://aws.amazon.com/elasticloadbalancing/
Not that it matters this far past Assignment 4, but one possible implementation of the performance monitoring my partner and I were wondering about was taking a derivative to see the change in number of requests over time. It didn't seem to work (or we didn't do it right), but I'm still convinced that this is a viable implementation. Are any performance monitors implemented that way?