Rensselaer Why not change the world? 1 Computational Center for Nanotechnology Innovations Rensselaer Why not change the world? A Computational and Research Center dedicated to Computational Nanotechnology Innovations A University/Industry/State Partnership Rensselaer Overview 3 Rensselaer Why not change the world? Educates the leaders of tomorrow for technologically based careers Schools – Architecture, Engineering, Humanities and Social Sciences, Management and Technology, Science 6,200 resident students – 5,000 undergraduate, 1,200 graduate Private institution founded in 1824 450 faculty, 1400 staff CCNI Vision - 1 Rensselaer Why not change the world? The Computational Center for Nanotechnology Innovations (CCNI) will bring together university and industry researchers to address the challenges facing the semiconductor industry as devices shrink in size to the nanometer range. 4 CCNI Vision - 2 Rensselaer Why not change the world? To account for the interactions of atoms and molecules up to the behavior of a complete device, the CCNI must develop a new generation of computational methods to support the virtual design of the next generation of devices which will require the massive computing capabilities of the CCNI. 5 CCNI Vision - 3 Rensselaer Why not change the world? The resulting virtual design methods will further expand New York State’s leadership position in nanotechnology. 6 Industry Needs Rensselaer Why not change the world? Needs – Technical and cost constraints are limiting the growth of the semiconductor industry and nanotechnology innovations – Computational nanotechnology is essential for decreasing the time from concept creation to commercialization 7 CCNI Goals Rensselaer Why not change the world? Goals Provide leadership in the development and application of computational nanotechnologies Establish partnership to create world class competencies on design-to-manufacturing research capabilities Produce new integrated predictive design tools for nano-scale materials, devices, and systems Spur economic growth in the Capital district, NYS & beyond 8 Facilities and Capabilities 9 Computational Systems – 100 teraflops of computing – Heterogeneous computing environment Rensselaer Technology Park – 4300 sq. ft. Machine Room – Business Offices – Systems and Operations support – Scientific Support Rensselaer Why not change the world? Layout of CCNI 10 Rensselaer Why not change the world? Partners to Build CCNI Rensselaer Why not change the world? Design and Engineering 11 Partners to Build CCNI Rensselaer Why not change the world? Architect 12 Partners to Build CCNI Rensselaer Why not change the world? Turner Construction Company 13 CCNI Construction Raised Floor 14 Rensselaer Why not change the world? CCNI Construction Cooling Towers 15 Rensselaer Why not change the world? CCNI Construction Lobby 16 Rensselaer Why not change the world? CCNI Installation Blue Gene Racks and Inter-rack Cables 17 Rensselaer Why not change the world? CCNI Installation Blue Gene Racks Without Covers 18 Rensselaer Why not change the world? CCNI Installation Rensselaer Why not change the world? Blade Racks, Storage Racks, and Network Cables 19 CCNI – Blue Gene/L Blue Gene/L System – – – – – 16 rack IBM Blue Gene/L system #7 on Top 500 Supercomputer List 32,768 PowerPC 700 MHz processors 12 TB of memory total Compute Node Kernel Simple, flat, fixed-size, address space Single threaded, no paging Familiar POSIX interface Basic file I/O operations – Two modes - coprocessor or virtual mode 20 Rensselaer Why not change the world? Blue Gene/L Hardware Optimized Communications 1) 3D Torus 2) Collective Network 3) Global Barrier/Interrupt 4) Gigabit Ethernet (I/O & connectivity) 5) Control (system boot, debug, monitoring) Rensselaer Why not change the world? 16 Racks 91.7 TF/s (peak) 12 TB 5.7 TF/s (peak) 512 GB or 1 TB 180 GF/s 16 or 32 GB 11.2 GF/s 2 GB 5.6 GF/s 4 MB 21 Blue Gene Architecture 22 Rensselaer Why not change the world? CCNI – Blade Servers Blade Server Cluster – – – – 462 IBM LS21 blades 1,848 Opteron 2.6 GHz cores 5.5 TB of memory total 4X InfiniBand interconnect (10 Gbps) – Red Hat Linux 23 Rensselaer Why not change the world? CCNI – Large Memory Rensselaer Why not change the world? AMD and Intel SMP Servers 40 IBM x3755 servers – Each with 8 Opteron 2.8 GHz cores and 64 GB of memory 2 IBM x3755 servers – Each with 8 Opteron 2.8 Ghz cores and 128 GB of memory 2 IBM x3950 servers – One with 64 Xeon 2.8 GHz cores and 128 GB of memory – One with 32 Xeon 2.8 GHz cores and 256 GB of memory All with 4X InfiniBand interconnect All with Red Hat Linux Power SMP Server – IBM p590 – 16 Power 5+ 2.1GHz processors – 256 GB of memory – AIX 24 CCNI – Disk Storage File Storage – Common file system for all hardware – IBM General Parallel File System, GPFS – 832 TB of raw disk storage – 52 IBM x3655 file server nodes – 26 IBM DS4200 storage controllers GPFS – High performance parallel I/O – Cache consistent shared access – Aggressive read ahead, write behind 25 Rensselaer Why not change the world? Rensselaer CCNI Networking Why not change the world? State International Local fiber: CCNI/Campus/NYSERNet 26 CCNI – Research Areas – Nanoelectronics modeling and simulation – Modeling of material structure and behavior – Modeling of complex flows – Computational biology – Biomechanical system modeling – Multiscale methods – Parallel simulation technologies 27 Rensselaer Why not change the world? Nanoelectronics Modeling and Simulation Rensselaer Why not change the world? – Functionality of new materials and devices – Fabrication modeling – Mechanics of nanoelectronic systems – Application to the design of new devices carbon nanotube T-junctions (Nayak) 28 submicron to nano (Huang) Modeling of Material Structure and Behavior Rensselaer Why not change the world? – Modeling and design of material systems – Modeling of energetic materials – Multiscale modeling of nanostructured polymer rheology Designed Interfaces Controlled Inhomogeneity Macroscopic Layering Matrix Compatible Block (MCB) Property Enhancing Block (PEB) Multiscale modeling of polymer rheolgy 29 Modeling of Complex Flows – Hierarchic modeling of turbulent flows – Modeling of biological systems flows 30 Rensselaer Why not change the world? Computational Biology Rensselaer Why not change the world? – Protein structure and interactions with small molecules – Membranes and membrane protein structure and function – Modeling cellular processes and communities of cells dppc cholesterol fentanyl 31 Biomechanical System Modeling – Virtual biological flow facility for patient specific surgical planning – Distributed digital surgery – Biomedical imaging via inverse problem construction 32 Rensselaer Why not change the world? Multiscale Science and Engineering – Multiscale mathematics and modeling – Adaptive simulation systems applied to applications 33 Rensselaer Why not change the world? Parallel Simulation Technologies Rensselaer Why not change the world? – High-performance network models – Optimistic parallel approaches – Multi-level parallel network models Geometric model 34 Partition model Initial mesh (1,595 tets) Partitioned mesh Adapted mesh (23,082,517 tets) Thus Begins the CCNI Odyssey 35 Rensselaer Why not change the world? Questions 36 Rensselaer Why not change the world? Rensselaer Why not change the world? 37