The Blue Gene/L supercomputer has been designed from the ground up with a focus on power/performance efficiency. The ultimate goal was to achieve extreme scalability and high application performance under the power and thermal constraints of existing data centers. To achieve this goal, emphasis was put on an integrated system solution. To ensure optimal system operation, Blue Gene/L is an integrated solution combining innovative system software, tools, architecture, system design, and packaging at all levels.

Blue Gene/L exploits application parallelism with an architecture that can scale to 64 racks, with a total of 65536 dual-processor computer nodes. Each compute node consists of a single compute ASIC integrating all system functions: integer and floating-point computing engines, multiple levels of cache, memory controller, and multiple network interconnects. With up to 131072 PowerPC processor cores in a single system, each PowerPC processor exploits data-level parallelism through a high-performance SIMD floating point unit.

To support good application scaling on such a massive system, special emphasis was put on efficient communication primitives by including five highly optimized communication networks. After an initial introduction of the Blue Gene/L system architecture, we analyze power/performance efficiency for the Blue Gene/L system using performance and power characteristics for the overall system performance. A high-performance software stack consisting of operating system services, compilers, message-passing infrastructure, high-performance computing libraries and middleware, debuggers, and performance tuning tools complete an integrated solution within a holistic design approach. In this tutorial, we will discuss both hardware and software design aspects of Blue Gene/L, with emphasis on the tradeoffs that were key to delivering a successful system.

This tutorial is directed at researchers and practitioners in the fields of hardware and software system architecture.

The Blue Gene/L project has been supported and partially funded by the Lawrence Livermore National Laboratories on behalf of the United States Department of Energy, under Lawrence Livermore National Laboratories Subcontract No. B517552.

Presenters

Valentina Salapura
IBM T.J. Watson Research Center, Yorktown Heights, NY

Valentina Salapura is a Research Staff Member with the IBM T.J. Watson Research Center. Dr. Salapura has been a technical leader for the Blue Gene program since its inception. She has contributed to the architecture and implementation of three generations of Blue Gene Systems focusing on multiprocessor interconnect and synchronization and multithreaded architecture design and evaluation. Dr. Salapura has been a leader in the definition and evaluation of the BlueGene architecture value proposition. Before joining IBM, Dr. Salapura was Assistant Professor with the Dept of Computer Engineering at Technische Universitt Wien, where she was a leading contributer to the architecture and implementation of the TTA time-triggered architecture and configurable microprocessor architectures. Dr. Salapura is the author of over 50 papers on processor architectures and high-performance computing, and holds many patents in this area. Dr. Salapura has served on the program committees of several international conferences and workshops, and is General Co-Chair for Computing Frontiers 2006. She received the Ph.D. degree from Technische Universitt Wien in 1996, and MS degrees in Electrical Engineering and Computer Science from University of Zagreb, Croatia.

Jose E. Moreira
IBM Systems and Technology Group, Rochester, MN, jmoreira@us.ibm.com

Jose E. Moreira received B.S. degrees in physics and electrical engineering in 1987 and an M.S. Jose degree in electrical engineering in 1990, all from the University of Sao Paulo, Brazil. He received his Ph.D. degree in electrical engineering from the University of Illinois at Urbana-Champaign in 1995. Since joining IBM in 1995, he has been involved in several high-performance computing projects, including the Teraflop-scale ASCI Blue-Pacific, ASCI White, and Blue Gene/L. Dr. Moreira was a manager at the IBM Thomas J. Watson Research Center from 2001 to 2004 and he is currently the Software Systems Architect for the IBM e-server Blue Gene solution. Dr. Moreira is the author of over 70 publications on high- performance computing. He has served in various thesis committees and has been the chair or vice-chair of several international conferences and workshops. Dr. Moreira is responsible for defining the technical characteristics of the IBM e-server Blue Gene solution, which started shipping to customers in the last quarter of 2004. Over 500 Teraflops of Blue Gene/L computing capacity have been installed at multiple sites in the US, Europe, and Japan. In his job, Dr. Moreira interacts closely with software developers, hardware developers, system installers, and customers to guarantee that the delivered systems work effectively and accomplish their intended missions successfully.