The use of grid computing in various research projects has yielded a number of important results in high-energy physics, body surface, molecular biology, astrophysics, medicine and has given impetus to the construction and use of modern computer infrastructures in research.
Wide adoption of grid computing in various research projects has yielded noticeable results in fields of high-energy physics, surface physics, molecular biology, astrophysics, medicine and has given momentum to development and use of modern computing infrastructures in scientific research.
Each research project relies on its own in-house specialized application software for modeling and data processing. Controlling the relevance of installed versions, configurations, available software libraries and porting of application software incompatible with software and hardware architecture of the grid require significant administrative efforts in the context of heterogeneous distributed infrastructure of the grid. For each virtual organization which is granted access to the computing resources, individual installation of libraries, application software and configuration of the runtime environment should be performed on each resource provider in the grid.
Use of virtual machines for creating isolated, manageable, independent runtime environments of grid services has become widespread in grid infrastructures. Thus, the crucial task is to develop methods for integrating virtualization technologies into the grid, which will enable running virtual machines as a job in the grid providing interactive access to application software inside. Such methods should allow the execution of application software in a controlled, isolated, computing environment of virtual machines running as grid jobs on the computing resources of grid-infrastructure independently of the compute cluster worker nodes environment. The combination of the possibility of creating an execution environment for the application software and interactive interaction with it in virtual machines running as grid jobs enables using more flavours of application software, which will significantly expand applications of grid technologies.
Objective of this paper is to developing methods and tools of creating computing environments that do not depend on the configuration of heterogeneous resources of grid infrastructure through integrating virtualization technologies and running application software in virtual machines running as grid jobs.
A new method of running virtual machines as a task grid has been developed, which allows use of container and hardware-assisted virtualization technologies. A new method of interaction with application software running in the grid has been developed, which enables interactive mode and use of graphical interface for different families of operating systems. A new method of executing application software in a virtual machine running as a grid job has been developed, which allows configuring the execution environment and running application software in batch and interactive modes of for different families of operating systems. The Rainbow software solution developed in the work implements methods of running a virtual machine as a grid job with interactive access to application software and allows the use of technologies of hardware-assisted virtualization and container virtualization. Developed architecture of grid service allows automating launch of virtual machines as grid jobs, manage virtual machine lifecycle, interacting with services of cooperation layer of the grid and providing means of integration into software packages of virtual organizations.
Developed techniques and architectural principles of the formation of an isolated, independent from working node of computing cluster of the grid infrastructure managed environment for running of the application software in the grid jobs. Mathematical modeling by means of transition systems and Petri Nets for modeling grid service of starting automation of virtual machine as grid tasks was conducted and it confirmed that such network is alive, and such mathematical tools can be applied to grid services.
