摘要
:
The paper presents an architectural framework and algorithms forengineering dynamic real-time distributed systems using commercialoff-the-shelf technologies. In the proposed architecture, a real-timesystem application is developed...
展开
The paper presents an architectural framework and algorithms forengineering dynamic real-time distributed systems using commercialoff-the-shelf technologies. In the proposed architecture, a real-timesystem application is developed in a general-purpose programminglanguage. Further, the architectural-level description of the systemsuch as composition and interconnections of application software andhardware, and the operational requirements of the system such astimeliness and survivability are specified in a system descriptionlanguage. The specification of the system is automatically translatedinto an intermediate representation (IR) that models the system in aplatform-independent manner. The IR is augmented with dynamicmeasurements of the system by a language runtime system to produce adynamic system model. The dynamic model is used by resource managementmiddleware strategies to perform resource management that achievestimeliness and survivability requirements. We present two classes ofalgorithms: predictive and availability-based, for performing resourceallocation. To validate the viability of the approach, we use areal-time benchmark application that functionally approximates dynamicreal-time command and control systems. The benchmark results illustratethat the middleware is able to achieve the desired timelinessrequirements during a number of load situations. Furthermore,availability-based allocation algorithms perform resource allocationless frequently, whereas predictive algorithms give a better steadystate performance for the application
收起