Mapping "heavy" scientific applications on a lightweight grid infrastructure
Lazar Kirchev, Minko Blyantov, Vasil Georgiev, Kiril Boyanov, Ian Taylor, Andrew Harrison, Stavros Isaiadis, Vladimir Getov, Natalia Currle-Linde · WestminsterResearch (University of Westminster) · 2006
An effective security and resource management is required for the provisioning and sharing of resources while keeping autonomy and stability of their environments.The models for security control and resource provisioning and sharing are to be implemented by closely-connected modules which interact tightly with information service and stand on role-based security model.These services are local to the cluster level but all of them have the ability to interact with the according neighbours in adjacent clusters.User, security, resource and information management have their representatives on global (grid) level and they are able to interact with the same services in cluster level in order to insure global centralized control.Global services are only viable when the system is used on global or extragrid level.In this paper we present the combined architecture of security and resource management and their close interaction with information service in a lightweight multilevel grid system based on highlevel middleware.It is important to check the performance and functionality parameters of the architecture at an early phase of architecture design.That is why here we consider the possibility for mapping of typical scientific application scenarios on several commodity computing clusters running our lightweight multilevel grid middleware.A set of representative use cases is listed in this paper along with taxonomy of the imposed requirements to the underlying infrastructure.For the purpose of exemplary mapping we choose an application scenario of molecular dynamics simulation which is a typical data-and computation-intensive asynchronous application.Here is given the technical mapping of this use case to the underlying infrastructure, paying particular attention to the possibility to implement high-level grid-unawareness for the use or application developer, to identify the compatibility of logical system and technological infrastructure used, as well as service and workflow representation.