| 摘要 |
Partitioning execution of a program between a client device and a cloud of network resources, exploits the asymmetry between the computational and storage resources of the cloud and the resources and proximity of the client access device to a user. Programs may be decomposed into work units. Those work units may be profiled to determine execution characteristics, modeled based on current state information and the profile, and a model performance metric (MPM) generated. Based on the MPM, work units may be partitioned between the client and the cloud. |
