See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/258580735

HPC in Abstract

Article · September 2013

CITATIONS READS
0 825

2 authors:

Hemprasad Y. Badgujar Ravindra Thool

Shri Guru Gobind Singhji Institute of Engineering and Technology Shri Guru Gobind Singhji Institute of Engineering and Technology
13 PUBLICATIONS 1 CITATION 72 PUBLICATIONS 951 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Person Re-Identification View project

MANET topology View project

All content following this page was uploaded by Hemprasad Y. Badgujar on 02 June 2014.

The user has requested enhancement of the downloaded file.

Cover Hemprasad Y Badgujar* and Dr. R C Thool**

Story Information Technology Department SGGS IE&T Nanded

HPC in Abstract
This is an era of High Performance processors having SISD architecture (first Clusters, Grids) having CPUs, MEMORY,
Computing abbreviated as HPC. HPC highly super scalar HPC system). Next GPUs, Multicore Processors, Arm
evolved due to increase in demand of decade of HPC was 1970s with vector Processor even DSPs and FPGA,
processing speed. It is a hybrid technology processors having SIMD architecture. Networking technologies (network
that continuously changes its colors, Later phase in 1980s was marked with topology, data transfer bandwidth &
so it is hard to define it in static words. massively parallel processors with MIMD latency, network protocols, network
However, if we still persist to define it the architecture having networked processors connections including infiniband, Ethernet
effort leads to a definition as below – with own memory and OS. Later in 1990s and proprietary connections with Fault
“HPC is a branch of computer science term ‘cluster’ was introduced for indicating Tolerance),
that concentrates on combining the connected standalone computers. These Software technologies, which include
power of computing devices, networking were mostly developed and designed suitable compilers for certain architecture,
infrastructure and storage systems by private communities for services of programming models (MPI, SHMEM,
using parallel hardware, networking and commodity hardware for networking. Grid PGAS etc.) and algorithms for parallel &
software technology that utilizes human was also introduced in the same decade distributed systems, use of middleware
expertise to gain optimized and sustained having foundation on collaboration of for bridging applications (Open-source,
performance for solving scientific, geographically distributed organizations freeware, commercial) and operating
engineering, big data and research and processing nodes for end users. Today system, and parallel programming for
problems including grand challenges like the cloud is one of the best sources of problemsolving methodologies
computer modeling, simulations and HPC through several massively scalable (Data parallelism, Pipelining, Functional
analysis of data.” and dynamic services as networking, Parallelism), programming languages (C/
HPC always focuses on developing processing, data storage, applications etc. C++, Python, Fortran etc. with parallel
high performance and it is achieved by All these services are provided through platforms and technologies like Open MP,
parallel processing systems, parallel Internet. If we need more information MPI Programming , UPC [Unified Parallel
programming algorithms and high about today’s fastest HPC systems, these C], OpenCL, CUDA, OpenCV, etc.)
bandwidth & low latency network by are listed in top500.org, out of which first If we go with the concept of HPC
incorporating computational and process 5 are shown in table 1. we can not avoid concepts of Moore’s
administration techniques of parallelism. As mentioned earlier HPC is hybrid law, concept of Von Neumann, Flynn’s
Since past several decades, HPC technology that covers a wide range of Taxiometry (SISD, SIMD, MISD, MIMD),
systems were only used for crunching technologies for example: topologies of computers, processors,
numbers for scientific research, Hardware technologies, which memory communication networks, data
engineering and big data problems. Now include computer architectures as New concurrency and correctness (data races,
they are open to everyone. Today HPC Generation processor architectures (IBM atomic operations, deadlocks, live lock
has almost become an essential part of CELL BE, Nvidia Tesla GPU, Intel Larrabee etc.), memory management (shared
businesses also for their daily procedures Microarchitecture and Intel Nehalem memory, distributed memory, semaphores
in order to help them become more microarchitecture ) and traditional and hybrid environments), partitioning,
competitive and profitable. processor architectures (Single Processor, data dependency, synchronization,
So far as history of HPC is concerned, SIMD, Vector computers, MPP, SMP, limits and cost measurement of parallel
it started in the decade of 1960s with scalar distributed systems, Constellations, programming (Speedup, Efficiency,

NAME SPECIFICATION CORES RMAX PFLOP/S POWER MW

Tianhe-2 NUDT, Intel Ivy Bridge (12C, 2.2 GHz) & Xeon Phi (57C, 3,120,000 33.9 17.8
(Milkyway-2) 1.1 GHz), Custom interconnect
Titan Cray XK7, Opteron 6274 (16C, 2.2 GHz) + Nvidia Kepler 560,640 17.6 8.3
(14C, .732 GHz), Custom interconnect
Sequoia IBM BlueGene/Q, Power BQC (16C, 1.60 GHz), Custom 1,572,864 17.2 7.9
interconnect
K computer Fujitsu SPARC64 VIIIfx (8C, 2.0GHz), Custom 705,024 10.5 12.7
interconnect
Mira IBM BlueGene/Q, Power BQC (16C, 1.60 GHz), Custom 786,432 8.16 3.95
interconnect
Table 1 : Top 5 super computer

CSI Communications | September 2013 | 10 www.csi-india.org

 Amdahl’s Law, Gustavson’s Law), Power- before designing HPC systems are - reservoir modeling, economic/financial
Aware Computing and Communication, Physical limitations of current hardware applications for risk analysis, portfolio
Processing Techniques, Memory & and technology such as speed of light, management & automated trading,
Interconnect Design and Software Power CPU heat dissipation etc., Economic plasma & nuclear physics, Quantum
Management. factors (to achieve good computing physics /chemistry, molecular biology /
Regarding HPC there are still many performance their is need of costlier dynamics, bio science & human genomes
questions, which arise in our mind – e.g. processing hardware but sometimes for drug discovery & diseases detection/
What is the need of HPC? Why do we need even cheaper components can be used to prevention. It also helps in business
this much processing power? Who are the achieve comparable levels of aggregate models in data warehousing, in LOB (Line
HPC users and who are the providers? performance), Scalability (allows data of Business) Applications, in Processing
How can we use and create HPC systems? to be subdivided into fragments for Transactions.
Which factors should be considered processing and to fit in algorithms So if you have a question in mind
before designing HPC systems? What and resources individually to increase as to – “Who wants HPC?” - the answer
are the benefits of HPC to the society and performance), Memory (ratio of aggregate is – All those want HPC, who want faster
businesses? memory bandwidth and processing power execution to provide solutions, quick
There is need of HPC where scenarios should increase together at a reasonable decisions, competitive products and quick
are too dangerous to be tested in real life cost ) and last but not the least expert models of simulation.
world and for scenarios that are very much man power. For those who want to adapt HPC,
time consuming. Some scenarios are An HPC system includes fast and high the question is - Where to start? There is
extremely complex to solve on common processing devices, mass data storage lot of material available on HPC concepts
systems. and collection management capacity, and terms. They are important to
There are a large number of scenarios optimized network with high transmission understand HPC. 1st step is learning HPC
that are far beyond the limitations speed. semantics and basics of technology. 2nd
faced by present day systems and tools. As HPC systems become more step is learning techniques of designing
Simulating, modeling and designing of affordable and accessible, there will and writing programs, 3rd step is trying to
these scenarios on HPC will gain improved be many areas where HPC will have implement / demonstrate programming
and quality testing with more number of significant presence like research module on computer systems. Take help
scenarios being tested, optimized time institutes, government labs, universities from open source programming samples
utilization and easier management even in and academic institutions for basic to improve and optimize utilization of
case of complex scenarios. HPC also helps & applied research. HPC systems are HPC hardware.
to open new windows, new possibilities of useful for weather forecasting for near The future of HPC will be achieving
discoveries and innovations for scientists, term & climate/earth modeling, defense Petascale Computing to Exascale
engineers and also for all to think of at a and energy research, mechanical / computing, transfer digital data
faster pace than ever before. So don’t be electrical design & drafting by 2D & 3D transfer for processing from Wired
afraid to let your mind wonder. design verifications, geo science & geo Communication Channel to Optical
Important factors to be considered engineering for oil, gas exploration & Communication Channels, Use of new
processor architecture, design of Quantum
Computers to increase clock speed and
As HPC systems become more aff ordable and accessible, there will transistor density in and for HPC systems,
be many areas where HPC will have signifi cant presence like research new continuous innovations and use of
institutes, government labs, universities and academic institutions for recent developments in Nanotechnology
basic & applied research. & Biotech and its impact on improving
processing and data storage systems.
n

Hemprasad Y Badgujar received his B.E. in Computer Engineering (University of Pune) in 2010, M-Tech in Information
Technology (SGGSIE&T Nanded Gov. Autonomous) in 2013. Presently he is working in NDMVP’s Karmaveer Adv.
Babural Ganpatrao Thakare College of Engineering Nashik. Areas of research interest are in High Performance
Computing, Parallel & GPU Computing, Computer and Network Security, Computer Vision, User Authentication &
Identification Systems

Prof. Dr. R C Thool received his B.E and M.E Electronics from SGGS Institute of Engineering and Technology,
About the Authors

Nanded (Marathwada University, Aurangabad) in 1986 and 1991 respectively. He obtained his Ph.D. in Electronics
and Computer Science from SRTMU, Nanded and is presently working as Professor in Department of Information
Technology at SGGS Institute of Engineering and Technology, Nanded. Areas of research interest are Computer
Vision, Image Processing and Pattern Recognition, Networking and Security, High Performance Computing, Parallel
& GPU Computing. He has published more than 50 papers in National and International Journals and Conferences.
He is Member of IEEE, ISTE, ASABE (USA) and also a member of University Co-ordination committee of Nvidia and
Expert on AICTE committee and NBA.

CSI Communications | September 2013 | 11

View publication stats