Afgan Enis, Baker Dannon, Nekrutenko Anton, Taylor James
Biology Department, Math & Computer Science Department, Emory University, Atlanta, GA.
Huck Institute for the Life Sciences, Penn State University, University Park, PA, USA.
Concurr Comput. 2012 Aug 25;24(12):1349-1361. doi: 10.1002/cpe.1836. Epub 2011 Aug 26.
Modern scientific research has been revolutionized by the availability of powerful and flexible computational infrastructure. Virtualization has made it possible to acquire computational resources on demand. Establishing and enabling use of these environments is essential, but their widespread adoption will only succeed if they are transparently usable. Requiring changes to applications being deployed or requiring users to change how they utilize those applications represent barriers to the infrastructure acceptance. The problem lies in the process of deploying applications so that they can take advantage of the elasticity of the environment and deliver it transparently to users. Here, we describe a reference model for deploying applications into virtualized environments. The model is rooted in the low-level components common to a range of virtualized environments and it describes how to compose those otherwise dispersed components into a coherent unit. Use of the model enables applications to be deployed into the new environment without any modifications, it imposes minimal overhead on management of the infrastructure required to run the application, and yields a set of higher-level services as a byproduct of the component organization and the underlying infrastructure. We provide a fully functional sample application deployment and implement a framework for managing the overall application deployment.
强大且灵活的计算基础设施的可用性彻底改变了现代科学研究。虚拟化使得按需获取计算资源成为可能。建立并启用这些环境至关重要,但只有在它们能够透明使用时,其广泛采用才能成功。要求对正在部署的应用程序进行更改,或者要求用户改变使用这些应用程序的方式,都构成了基础设施被接受的障碍。问题在于应用程序的部署过程,以便它们能够利用环境的弹性并将其透明地交付给用户。在此,我们描述了一种将应用程序部署到虚拟化环境中的参考模型。该模型基于一系列虚拟化环境共有的底层组件,它描述了如何将那些原本分散的组件组合成一个连贯的单元。使用该模型能够在不进行任何修改的情况下将应用程序部署到新环境中,它对运行应用程序所需的基础设施管理施加的开销最小,并作为组件组织和底层基础设施的副产品产生一组更高级别的服务。我们提供了一个功能齐全的示例应用程序部署,并实现了一个用于管理整体应用程序部署的框架。