Sustainable Adaptive Grid Supercomputing: Multiscale Simulation of Semiconductor Processing across the Pacific

Abstract

We propose a reservation-based sustainable adaptive Grid supercomputing paradigm to enable tightly coupled computations of considerable scale (involving over 1,000 processors) and duration (over tens of continuous days) on a Grid of geographically distributed parallel supercomputers. The paradigm is demonstrated for an adaptive multiscale simulation application, in which accurate but compute-intensive quantum mechanical (QM) simulations are embedded within a classical molecular dynamics (MD) simulation only when and where high fidelity is required. Key technical innovations include: 1) an embedded divide-and-conquer algorithmic framework to maximally expose data and computation localities for enhanced scalability; 2) a buffered-cluster hybridization scheme to adaptively adjust MD/QM boundaries to maintain the model accuracy; and 3) a hybrid Grid remote procedure call (GridRPC) + message passing interface (MPI) Grid application framework to combine flexibility (adaptive resource allocation and migration), fault tolerance (automated fault recovery), and efficiency (scalable management of large computing resources). We have achieved an automated execution of multiscale MD/QM simulation on a Grid consisting of 6 supercomputer centers in Japan and the US (in total of 150 thousand processorhours) for the dynamic simulation of implanted oxygen atoms in a silicon substrate, in which the number of processors changes dynamically on demand and resources are allocated and migrated dynamically according to both reservations and unexpected faults. The simulation results reveal a strong dependence of the oxygen penetration depth on the incident oxygen-beam position, which is useful information to further advance SIMOX (separation by implanted oxygen) technique to fabricate high speed and low power-consumption semiconductor devices.

De Novo Ultrascale Atomistic Simulations On High-End Parallel Supercomputers

“…However, the solution is sensitive to the choice of the termination atoms, and thus the domains need to be determined manually before the simulation . The buffer-layer approach of the EDC-DFT algorithm considerably reduces this sensitivity, and accordingly we are among the first to automate the adaptive domain redefinition during simulations (Takemiya et al 2006).…”

“…Our current adaptive hierarchical simulation manages the error based on a simple heuristic, i.e. the deviation of bond lengths from their equilibrium values, for which the MD interatomic potential has been trained (Takemiya et al 2006). For tighter error management, we extend this approach by encoding the deviation of local topological structures of atoms from those in the training set, based on an abstraction of the structures as a graph and its shortest-path circuit analysis.…”

The Grid Enablement and Sustainable Simulation of Multiscale Physics Applications

“…Our group has also proposed a hybrid MPI+GridRPC programming model to combine the advantages of both MPI and GridRPC. Using this method, we performed some large-scale experiments over TeraGrid, but most of the work such as resource reservation and error recovering was done manually [12]. We thus have developed an integrated framework [5] to provide users with APIs to develop fault-tolerant, resource-aware, and selfload-balancing applications.…”

“…Different from other applications in time and scale, this work enables a physics application to run in a crosscontinent grid environemnt on top of our framework which makes it uniquely different from previous work [12]. In addition, this work focuses on the application perspective, while our another work [5] discusses the middleware perspective.…”

EA-based Problem Solving Environment over the GRID

“…This unique parallelization technique employing GridRPC and GridMPI was first used in (Takemiya et al, 2006) for a specific problem. MHGrid deploys this technique as a general model for dynamic grain size definition.…”