Pattern description

This pattern essentially has two different parts handling failover and handling load burst. Failover is the ability to switch over to another "ready" system when an abnormal termination occurs. This is usually automatic when the system detects the termination and assigns control to the ready system. One of the key goals when designing failover is to make this transparent to the end users. Users might need to reconnect, but application state or data may be centralized so that the "ready" system can pick up where the failed one ended. Deciding capacity requirements for a system is again a systematic and scientific computation process —different metrics such as the number of transactions, required response time, and the number of concurrent users may be used. In the end, a decision needs to be made and this number would be higher than the average of the metrics but lower than the occasional peak. A load burst is a sudden increase or spike in workload possibly just for...