Supercomputers Molecular Dynamics: Components And Architecture

Major components of Supercomputers

Discuss about the Supercomputers Molecular Dynamics.

A supercomputer can be defined as the computer with very high level of computational capacity when compared with the general purpose computer. The term supercomputer was coined in 1929 at New York for a set of large customized tabulators (Chen et al., 2014). The set of tabulators was made by IBM to support the activities in Columbia University. However, the supercomputers were introduced in the late 1960s. Seymour Cray working at Control Data Corporation built it. The early supercomputers used only few processors. The development of the supercomputers made it evitable for incorporating thousands of processors in the end of 1990s. The count of processors reached to tens of thousands in the multiple parallel supercomputers developed at the end of 20^th Century (Shaw et al., 2014,). The performance of the supercomputers is measured in floating point operations per seconds or FLOPS. The latest supercomputers have performance capacity of up to quadrillions of FLOPS. The following assignment has been made for analysis of the supercomputers and their structure and benefits. The study would help in evaluating the various components of the supercomputers and description of any one of them. The assignment would also highlight some innovative trends that may be possibly seen in the near future of supercomputers.

The supercomputers are very useful for solving highly calculative intensive tasks that have complex structure and operations. These complex operations include simulation, modeling, analyzing various types of data, cryptographic analysis, and intricate designing (Sridharan et al., 2013). The operations of supercomputers are helpful for weather forecasting, simulations of molecules behavior, and modeling of airplanes within wind tunnels. The analysis of data finds its use for the security agencies, astronomical data manipulation, and sequencing of the genome. The architecture of supercomputer consists of processors, memory, input and output devices, and interconnection.

Processors: The processors used in supercomputers are completely made up of pure silicon. The pure silicon consists of 99.99% silicon crystal and it is more costly when it is compared to gold in weight to weight (Liao et al., 2014). The size of silicon crystals are approximately 300 square meters. The processors consist of doping materials that helps in formation of wires, resistors, transistors, and capacitors. The components of doping materials also consist of rare elements of earth and their quantity is in minute volumes.

Memory: The memory used in the supercomputers is distributed in nature. Each processor of the supercomputer has been connected to an exclusive local memory with an network interface node ‘NI’ (Hord, 2013). The dedicated communication network is useful for connecting the nodes of the supercomputer. NORMA or no remote memory access is implied in the system for forming no global cache coherent shared memory address.

Discussion of Distributed memory system of Supercomputers

Input Output devices: The input output devices are helpful for inputting the devices into the system and making allowance for the output of the value. The input and output functions of the supercomputers have same concept of operations like in general computers. It includes devices for entering the values/data in the system for processing and devices used for displaying the output of the processed data (Bartolini et al., 2014). The various input devices like keyboards, reader, or microphone would act as the input devices for the supercomputers. The output devices mainly include the display monitor and printers.

Interconnections: The interconnection consists of a network subsystem in the supercomputer for targeting memory intensive applications of supercomputers (Anwar-Mohamed et al., 2014). The supercomputers are integrated with many nodes for the high performance of the network operations. The interconnection design of the supercomputer is based on the 64X64 high radix crossbar switches. The design of the interconnection depends upon the traffic amount, communication pattern, and physical constraints. However, there are probabilities of finding the bottleneck of the communication design for further enhancement and improvement of the performance.

The supercomputers are designed for scaling up of the entire system along with the computer performance (Rohr et al., 2014). The design space exploration of the computer devices is responsible for the evaluation of the feasibility of the networks. The complex architecture of the supercomputers makes them very complex from other standard computers. The characteristics differences between the standard computer and the supercomputers are specialization, controlled environment, operating system, hierarchy of memory designs, and programming language.

Distributed storage memory is very crucial in the structure of supercomputer. The parallel algorithm has been implemented in the structure of supercomputers. The processes by which the parallel algorithm can be implied in the structure of the supercomputers are 1 shared memory systems (uniform memory access UMA and cache coherent non uniform memory access ccNUMA) and 2 distributed memory systems (massive parallel processing and cluster computing) with each of them having a standardized programming model (Stephan & Docter, 2015). The memory system of the supercomputer is of hybrid structure with shared and distributed architectures. The memory system shares the same address space or memory at low level of computation nodes. Numerous storage appliances are used for forming the backbone of the new storage system that would be capable for the storing, ingesting, distributing, and processing of the research data at a rate of one terabyte per second.

The use of flash drive has made the data processing more easy and comfortable (Zgurskaya & Smith, 2016). The processing of the data has been largely improved and it has resulted in formation of the faster throughput of the data. The solid state devices used in supercomputer are prone to store the data on the flash memory chips instead of the magnetic platters that were used in hard drives. The solid state devices do not require any movement and it has made the system less rugged and vulnerable for the failure. The devices of solid state do not consume much energy and are classified as less power hungry devices. The speed of data processing is almost similar to ten times of the computer system of NAS system. The storage system comprises of 40 Peta-bytes of unprocessed data storage that can hold almost about 227000 miles of the books stacked to each other (Rohr, Neskovic & Lindenstruth, 2015). The performance of the supercomputers is measured in floating point operations per seconds or FLOPS. The latest supercomputers have performance capacity of up to quadrillions of FLOPS.

The supercomputers have incomparable increased value of the computational efficiency that can deliver more optimized performance with predictive models of the computerized storage. The 40 Peta bytes of storage are comparable to 40,960 computers with 1 Tera bytes of storage each (Ellingson et al., 2014). The supercomputers are built with high solid state drives for memory system. It would help in solving the problems of the computation faster than the systems that incorporates the traditional hard drives. The supercomputers are designed for scaling up of the entire system along with the computer performance. The memory used in the supercomputers is distributed in nature. Each processor of the supercomputer has been connected to an exclusive local memory with an network interface node ‘NI’. The dedicated communication network is useful for connecting the nodes of the supercomputer. The complex architecture of the supercomputers makes them very complex from other standard computers (Zgurskaya & Smith, 2016). The characteristics differences between the standard computer and the supercomputers are specialization, controlled environment, operating system, hierarchy of memory designs, and programming language.

The development of technology has been increased and it has seen abrupt pace of increased potential. The technology that has been largely innovated since the last decade is the Supercomputer technology (Stephan & Docter, 2015). The supercomputer technology has seen wide development of the implementation models. Some future trends of technology that can be used in supercomputers include:

Increment of power and performance: The Moore’s law has presumed that the technological development of the integrated chips has seen large increment in power generation. It can be predominantly said that after a span of 24 months, the overall capacity of the computer chips would be doubled (Anwar-Mohamed et al., 2014). The increment of the power of processors would result in increasing the processing speed and efficiency of the chips. The use of improved ICs for the CPUs of the supercomputers would help in increasing the overall processing efficiency of the device.

Deployment of the cloud function: The latest trend in storage and memory is the application of mobile cloud services (Hord, 2013). However, the cloud server can provide the space up to 10GB. The development of technology would be helpful for increasing this limit and integrate the cloud system with the operations of supercomputers.

Conclusion

The completion of the assignment had helped in realizing that the supercomputers have developed since the last three decades and the technology had made it evitable for incorporating tens of thousands of processors in the supercomputers. The latest supercomputers have a performance capacity of up to quadrillions of FLOPS. The supercomputers had achieved the qualities of solving highly calculative intensive tasks with complex structured operations. Some of the complex operations include simulation, modeling, analyzing various types of data, cryptographic analysis, and intricate designing that had been widely used for weather forecasting, simulations of molecules behavior, and modeling of airplanes within wind tunnels.

References

Anwar-Mohamed, A., Barakat, K. H., Bhat, R., Noskov, S. Y., Tyrrell, D. L., Tuszynski, J. A., & Houghton, M. (2014). A human ether-á-go-go-related (hERG) ion channel atomistic model generated by long supercomputer molecular dynamics simulations and its use in predicting drug cardiotoxicity. Toxicology letters, 230(3), 382-392.

Bartolini, A., Cacciari, M., Cavazzoni, C., Tecchiolli, G., & Benini, L. (2014, March). Unveiling Eurora—Thermal and power characterization of the most energy-efficient supercomputer in the world. In Design, Automation and Test in Europe Conference and Exhibition (DATE), 2014 (pp. 1-6). IEEE.

Chen, Y., Luo, T., Liu, S., Zhang, S., He, L., Wang, J., … & Temam, O. (2014, December). Dadiannao: A machine-learning supercomputer. In Proceedings of the 47th Annual IEEE/ACM International Symposium on Microarchitecture (pp. 609-622). IEEE Computer Society.

Ellingson, S. R., Dakshanamurthy, S., Brown, M., Smith, J. C., & Baudry, J. (2014). Accelerating virtual high?throughput ligand docking: current technology and case study on a petascale supercomputer. Concurrency and Computation: Practice and Experience, 26(6), 1268-1277.

Hord, R. M. (2013). The Illiac IV: the first supercomputer. Springer Science & Business Media.

Liao, X., Xiao, L., Yang, C., & Lu, Y. (2014). MilkyWay-2 supercomputer: system and application. Frontiers of Computer Science, 8(3), 345-356.

Rohr, D., Kalcher, S., Bach, M., Alaqeeliy, A. A., Alzaidy, H. M., Eschweiler, D., … & Alqwaizy, I. (2014, August). An energy-efficient multi-GPU supercomputer. In High Performance Computing and Communications, 2014 IEEE 6th Intl Symp on Cyberspace Safety and Security, 2014 IEEE 11th Intl Conf on Embedded Software and Syst (HPCC, CSS, ICESS), 2014 IEEE Intl Conf on (pp. 42-45). IEEE.

Rohr, D., Neskovic, G., & Lindenstruth, V. (2015). The L-CSC cluster: Optimizing power efficiency to become the greenest supercomputer in the world in the Green500 list of November 2014. Supercomputing frontiers and innovations, 2(3), 41-48.

Shaw, D. E., Grossman, J. P., Bank, J. A., Batson, B., Butts, J. A., Chao, J. C., … & Forte, A. (2014, November). Anton 2: raising the bar for performance and programmability in a special-purpose molecular dynamics supercomputer. In Proceedings of the international conference for high performance computing, networking, storage and analysis (pp. 41-53). IEEE Press.

Sridharan, V., Stearley, J., DeBardeleben, N., Blanchard, S., & Gurumurthi, S. (2013, November). Feng shui of supercomputer memory positional effects in DRAM and SRAM faults. In High Performance Computing, Networking, Storage and Analysis (SC), 2013 International Conference for (pp. 1-11). IEEE.

Stephan, M., & Docter, J. (2015). JUQUEEN: IBM Blue Gene/Q® Supercomputer system at the Jülich supercomputing centre. Journal of large-scale research facilities JLSRF, 1, 1.

Zgurskaya, H., & Smith, J. (2016). Supercomputer Simulations Help Develop New Approach to Fight Antibiotic Resistance. ORNL (Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States)).