As modern computer system advances, more and more distributed applications with multiple QoS requirements are emerging, such as waste water and natural water monitoring and control. It has become an urgent need to develop techniques that guarantee applications' QoS constraints and decrease the complexity of implementing such systems. Through this project, we have developed a new paradigm that utilizes the separation of concern design principle to simplify the design and development of distributed real-time applications. In particular, the developed design and implementation paradigm separates coordination among distributed objects from the functionalities of the objects themselves. We have also developed a set of effective methods and techniques to maintain deployed real-time application's QoS requirements.

We have addressed the challenge of developing and implementing distributed real-time applications in open environment from both system implementation simplification and multi-QoS guarantee perspectives. For system implementation simplification, we have developed a hierarchical Actor-Role-Coordinator (ARC) programming model to simplify the system design and implementation. We have also developed a concept lattice based event model for high level semantic reasoning and a eUML-ARC model for concurrency programming to further reduce the system implementation complexity and improve development efficiency. For multi-dimensional QoS guarantees, we have developed a similarity based measurement to quantify the satisfaction of multi-dimensional QoS requirements. We have also developed information hiding and voting algorithms to guarantee system's availability, reliability in open distributed environment.  In addition,  a profit-penalty aware scheduling algorithm to maximize system profit while guaranteeing application's QoS requirements, a minimal slack time and minimal distance scheduling algorithm to minimize the cost for guaranteeing real-time application's timing constraint in cloud environment have been developed. 

In addition to system level implementation simplification and scheduling algorithm development, we have also addressed the QoS guaranteeing problem from considering resource performance change perspective.  Traditionally, system resources are considered as constant and they do not change over their life time.  Such assumptions are not valid for real-time applications which often operate for long time.  In fact, resource performance degrades overtime, which is evidenced by all computer users' own experiences --- our computers get slower if they have been running for long time without a reboot.  Software aging is identified as one of the main causes of resource performance degradation, software rejuvenation is a commonly used effective technique to combat performance degradations.  We have developed a novel resource model, i.e., the $P^2$ resource model, that takes into consideration of performance-degradation and periodic-rejuvenation.  The $P^2$ resource model generalizes traditional resource models existed in the literature. Based on the $P^2$ resource model, we have developed a set of theory and scheduling algorithm to ensure real-time guarantees for real-time applications. 

We have also studied large amount of virtual machine launch data obtained through collaboration with Fermilab on FermiCloud.  The analysis of these data indicates that there are large variations when launching a virtual machine both  on a private and a public cloud.  Based on over  three months  data collected on both private and public clouds, we have developed a cloud resource performance variation model that allows us to predict the resource startup overhead.  The reference model is a small but an important step forward to pave the road for deploying real-time app...

Please report errors in award information by writing to: awardsearch@nsf.gov.