Improved load management based on resources reservation in distributed embedded systems

Load balancing strategy has been founded and applied widely as an effective measure of performance optimization for most conventional distributed systems. However, for the sake of distributed embedded system's inherent properties, especially for real-time property, load balancing strategy may not be desirable. On consideration of those properties, an improved load management scheme is presented based on a new way of resource reservation, which has considerable enhancements in satisfying real-time requirements. First, the conception of satisfaction to real-time requirement of incoming task is introduced in a new aspect. Then the implementation of resource reservation is narrated in detail. Under this concept of load management, the new incoming tasks are allocated to the most heavily loaded node, if only the service node can provide real-time satisfaction. Thus, a great amount of calculating resources is reserved for those tasks that require almost full resource of a service node. In addition, a central control model based on this improved load management is introduced into application and implement in building up a prototype system according to this model. During this simulation, the new load management strategy based on resource reservation shows great advantage according to real-time criteria against load balancing. The conclusion comes into being that the resource reservation strategy arising in the paper benefits from a distributed embedded system with high real-time requirements.