Enabling Time-sensitive Applications on Virtualized Computing Systems

Project Information

Discipline

Please Select... 

Orientation

Research 

The objective of this project is to address the research challenges for enforcing timeliness guarantees on virtual machines (VM) and efficient timeliness-driven resource management of virtualized systems, supporting scalable and flexible computing of applications with different degrees of timing constraints. This project proposes a new virtualized time-sensitive computing approach based on cooperative cross-host-guest scheduling, timeliness-driven VM management, and autonomic cloud-scale resource optimization.

System virtualization supports flexible resource sharing with strong isolation and convenient application deployment on customized execution environments. It is at the core of public and private cloud computing systems which can elastically provision resources on demand. Many applications can benefit from computing on virtualized systems, including those that are time sensitive, such as robotic control (which functions only if finished within a specific deadline) and visualization (in which a response-time violation means loss of interactivity). However, it is challenging for virtualized systems to deliver application-desired timeliness. First, typical VM architectures are designed for general-purpose usage but not tailored for responsive, predictable application executions. Second, the high resource consolidation enabled by virtualization also creates complex, dynamic resource contention, while time-sensitive applications are particularly vulnerable to variations of resource availability. Third, the resource management on a virtualized system is often optimized for fair sharing and overall throughput, but not for various application-specific timing constraints. This project will design novel solutions to address these limitations and enable virtualized time-sensitive computing.

This project will enable a broad range of time-sensitive applications on virtualized computing systems, thereby generating impact to potentially many areas where the use of virtualization and cloud has value but current technologies do not provide adequate support for time-sensitive workloads. By hosting such applications on customized VMs, it becomes possible for them to be conveniently deployed onto heterogeneous platforms and to efficiently harness available resources. By delivering different degrees of timeliness to different workloads, it becomes possible to support a wide variety of applications.

Project Contact

Project Lead

Ming Zhao (ming) 

Project Manager

Ming Zhao (ming) 

Resource Requirements

Hardware System

• foxtrot (IBM iDataPlex at UF)

 

We plan to use FutureGrid resources to conduct large-scale experiments in order to evaluate the time-sensitive cloud computing techniques developed in this project.

100 to 200 VMs for one for two months.

Project Timeline

Submitted

04/04/2014 - 23:25