The Design Of Distributed Real-time Task Scheduling Service

With advantages such as cross-platform, language-independency, and practicability, CORBA (Common Object Request Architecture) , the mainstream of DOC (Distributed Object Computing), has been widely used, making it the de facto industrial standard.CORBA-based distributed real-time system is an important kind of distributed application. This kind of application must respond to external event in time, finish the process of distributed information, and activate the distributed execution mechanisms in real time. The traditional single-computer-oriented real-time task scheduling mechanism can not satisfy the requirements of Distributed Real-Time Object Computing Environment (abbreviated as DROCE in this paper). In order to guarantee the real-time constraints and predictability of distributed real-time object applications, we must conduct research on the scheduling problem in DROCE.Based on the analyses of scheduling strategy and scheduling framework of TAO (The ACE ORB), this paper studies the real-time scheduling and management of global priory in DROCE: how to determine the global priory of the DRO (Distributed Real-Time Object) according to its timeliness and level of importance; logical presentation and transparent transfer of the global priory; the consistent mapping from global priory to local server priory; the prevention of priory inversion, etc.. With hard real time application as the application background, this paper presents the design and implementation of scheduling APIs which support scheduling strategy.This paper provides three main contributions to the design and implementation of scheduling service in DROCE:First, this paper comprehensively analyzes the scheduling service framework of TAO. It discusses the implementation of the TAO scheduling service and its various priority-processing details.Second, this paper conducts in-depth research on many distributed scheduling strategies. This paper points out the applicable environments for different kinds of scheduling strategies.Finally, based on the analysis to TAO scheduling service, this paper presents the design and implementation of scheduling APIs which support scheduling strategy (for WindowsNT/2000 and Sun Solaris) .Distributed real time scheduling service is the basis of distributed real time resource management technique. User's MQoS requirements will all be ultimately mapped into the priorities of the threads (processes) which execute the operations. This paper makes some contribution to the research of resource management technique under distributed real time object computing environment.