| 摘要 |
In one aspect of the invention, a method of dynamically switching among a plurality of fault tolerance schemes is provided. The fault tolerance schemes are associated with a fault tolerance mechanism that executes in a distributed system. The method comprises obtaining a wait time of at least one user interface event occurring in the distributed system. The wait time includes at least one of a communications time, a service time and a fault tolerance time. The method further comprises determining whether a mean of the wait time is greater than a predetermined mean wait time threshold. The method also comprises determining whether the communications time, the service time and the fault tolerance time are mutually independent when the mean of the wait time is greater than the predetermined mean wait time threshold. In addition, the method comprises determining whether the mean of the wait time can be improved by reducing a mean of the fault tolerance time when the communications time, the service time and the fault tolerance time are mutually independent. The method also comprises switching from a first fault tolerance scheme to a second fault tolerance scheme when the wait time can be improved by reducing the mean of the fault tolerance time.
|
