VLab is an interdisciplinary initiative dedicated to the development and promotion of methodology, tools, and knowledge in planetary materials science. It embodies an ambitious cyberinfrastructure (CI) development effort to enable complex studies. The CI includes a set of collaborative empowering tools to leverage the outstanding throughput of distributed resources for �first principles� computations of materials properties. VLab project is developing a cyberinfrastructure to enable execution of very complex calculations - mostly parameter sampling workflows - of materials properties composed by hundreds of tasks. The calculation of thermal properties of materials or elastic constants at high pressures and temperatures is an outstanding example. The VLab CI is implemented as a service oriented architecture (SOA), plus a portal responsible for the user interface. First principles calculations are carried out with the open source Quantum ESPRESSO package. Moreover, the ability to fully abstract the complexity and heterogeneity of the underlying distributed computational resources is an essential requirement. Computational studies requiring sampling of points in parameter space appear in several scientific fields such as environmental sciences, astrophysics, particle physics, computational fluid dynamics, and materials science, the subject of our special interest. Usually these studies comprise a large number of tasks, each one responsible for a given parameter domain. Fortunately, calculations related to different parameter domains are decoupled. This enables them to be executed concurrently. Moreover, coarse grained scientific applications can be efficiently executed on Grid resources, like those available on TeraGrid. Calculating thermodynamic properties and elastic constants of Earth forming materials at high pressures and temperatures from first principles is an important application in materials science. Calculation of this kind are structured as a multi-phase workflow, each phase requiring extensive parameter sampling. Cardinality of parameter point sets can exceed 10^4, with each point spanning a single compute-intensive task, i.e., a full fledged DFT based ab initio calculation, requiring as much as 10^16 floating point operations. Distributed computing, particularly Grid computing, is enjoying a growing acceptance in computational physics. The VLab project is developing a novel service oriented architecture (SOA) designed to be a facilitator in performing this kind of calculations, among others. The SOA is required to manage tasks in a distributed environment, leveraging the high throughput of distributed resources, in addition to the high performance of its constituent hosts. Additionally, it provide tools, interfaced to the user through ease of use portal, to manage execution, monitoring, and steering of workflows, data analysis, and visualization. VLab is capable of managing its own pool of resources. Nevertheless, we have developed a service that acts as an interface between VLab and Teragrid metaschedulers, mostly CondorG, enabling user access to the outstanding TeraGrid's computational resources. This way, VLab is enabled to act as a sofisticated TeraGrid Science Gateway, empowering users to execute jobs ranging from a single instance of a given ESPRESSO packege tool to a very complex workflow, with full abstraction of the complexity of the underlying grid.

wentzcov@cems.umn.edu,cesards@msi.umn.edu,pedros@msi.umn.edu

http://www.vlab.msi.umn.edu/

cyoun

Researchers should cite this work as follows:

• (2024), "VLab - Virtual Laboratory for Earth and Planetary Materials," https://sciencegateways.org/resources/vlab-virtuallaboratoryforearthandplanetarymaterials.

  BibTex | EndNote

1. ACCESS-CI
2. materials
3. Quantum ESPRESSO
4. SOA
5. VLab