Supercomputer is that computer with the help of which scientific calculations of top level can be done easily. Normal computer not able to perform these calculations. Supercomputer is used in advance research area.
FLOPS (Floating-point Operations Per Second) is the performance measure of Supercomputer, rather than MIPS (Million Instructions Per Second). There are some supercomputers also that can operate billion of billion FLOPS. Operating System of all Supercomputers since November 2017, is Linux.
Windows operating supercomputers out of list since 2015. ‘TOP500’ ranks the supercomputer and provide list twice a year. Some countries that perform Advance Research like U.S, China, Japan, European, Taiwan etc. always try to develop more powerful and technically superior supercomputer.
Uses of Supercomputer:
Main use of supercomputer is in technology and in performing difficult scientific calculations.
Some other uses of supercomputer are:
- Weather Forecasting
- Atomic and Molecular Modelling
- Physical Simulation
- In extraction of Oil & gas
- Climate Research
Use of Supercomputers was started in 1960’s, Cray was famous at that time. In supercomputing field US was leader for a long time. Japan worked purposefully in the field in 1980s and 1990s.
China was also in competition at that time. As stated above ‘TOP500’ presents its list of top 500 supercomputers twice a year. First one in June and the Second one in every November. Latest list of ‘TOP500’ is of November 2018. According to which we are going to tell about Best 10 supercomputers of the World.
Discussion on top of list:
China is the leading Country with 229 supercomputers listed in the List, whereas US is on the 2nd position with 108 supercomputers in the list. For the Comparison of best 10 computers LINPAC benchmark is used. Summit a supercomputer of US gets 143.5 PFLOPS which is the Best Supercomputer in the World, followed by Sierra another supercomputer of US ranked 2nd.
Now here we will distinguish Best 10 supercomputers on the basis of performance, Model, processor and operating system.
Distribution of supercomputers in the list of TOP500 according to Country:
Country No. of supercomputers
United States 108
United Kingdom 20
South Korea 6
Saudi Arabia 3
South Africa 2
1. SUMMIT (UNITED STATES)
Summit is a supercomputer of US. OLCF-4 was developed by IBM. Summit is use At Oak Ridge Laboratory. In ‘TOP500’ Summit top the list in November 2018 with capacity of 200 PFLOPS.
Current LINPACK benchmark of Summit is 143.5 petaflops. This Supercomputer is 3rd most energy adequate with power efficiency of 14.67 GLOPS/watt. Summit have reached exaops i.e. exa operations per second. Summit mainly will used for various purpose like cosmology, medicine and climatology.
Design of Summit:
A total comprehensible memory nearly 600 Giga Byte. 6*16= 96 giga byte High Bandwidth Memory + 2*8*32 = 512 giga byte Double Data Rate 4 Synchronous Dynamic Random-Access Memory that is admissible to all CPU and GPU. Additional 800 GB is used for storage purpose.
2. SIERRA (UNITED STATES)
Sierra is 2nd best supercomputer of US in the world according to ‘TOP500’. Sierra is built for Lawrence Livermore National Laboratory. This Supercomputer will use by NNSA (National Nuclear Security Administration). Sierra is used for Nuclear weapons of United States. Sierra is also built by IBM. This Supercomputer gives surety of safety, liability and effectiveness of weapons.
Design of Sierra:
Sierra is almost same as Summit Supercomputer in design. Sierra supercomputer use IBM POWER9 CPUS in addition with Nvidia Tesla V100 GPU.
3. SUNWAY TAIHULIGHT (CHINA)
Sunway TaihuLight is a supercomputer of China. According to November 2018, Sunway ranked 3rd best supercomputer of the world. LINPACK benchmark for this supercomputer was 93 PFLOPS. Taihu Light name is derived from Lake Taihu, Divine power.
This supercomputer is much faster than Tianhe-2. In Green500 this computer is ranked as 16th most energy efficient computer with power 6.1 GFLOPS/watt. Sunway was developed by NRCPC (National Research Center of Parallel Computer Engineering and Technology). Sunway is situated at National Supercomputing Center in Wuxi (China).
According to ‘TOP500’ for two years (June 2016 to June 2018) Sunway TaihuLight was world’s superfast computer. But in November 2018 this was defeated by Summit of IBM (US).
Design of Sunway TaihuLight:
Sunway TaihuLight runs on its own operating system SW26010 manycore 64-bit RISC processor. 256 processing cores are contained by every processor and for Management 4 auxiliary cores are provided.
Sunway TaihuLight supercomputer have 64 KB of memory for data and 16 KB memory for instructions. System communicate through chip system rather than old cache hierarchy.
4. TIANHE-2A (CHINA)
It is the supercomputer of China. Tianhe-2 meaning is Milky Way 2. Tianhe-2 has capacity 0f 33.86 PFLOPS. It is situated in NSC (National supercomputer Center) in Guangzhou in China.
It was said that this supercomputer was established with the help of 1300 engineers and scientist. In the list of ‘TOP500’ in June 2013 & November 2013, June 2014 & November 2014 and June 2015 & November 2015 this was the world’s fastest computer.
China plans to increase the capacity of supercomputer two times but US government didn’t permit to do so. That why China develop Sunway TaihuLight. China use its own operating system and ranks 4 in the list. Tianhe is meant to use for government security purpose. It is said that Tianhe-2A is quite difficult in use.
The location of Tianhe-2A is in Southern China where Temperature is warm than Northern China, so the electric consumption is 10% more in that area. This is the drawback of this supercomputer. Some Scientist take so much time to learn codes of this supercomputer that’s why it is difficult in use.
Design of Tianhe-2A:
China use its own operating processor. Tianhe-2A. This supercomputer has 16000 nodes, every node has two intel Ivy Bridge Xenon Processor and 3 Xenon Phi Coprocessor chips. Total memory possessed by nodes is 88 GB. Total memory of system was 1375 TB. System operates on Linux (Kylin) setup by NUDT.
5. PIZ- DAINT (SWITZERLAND)
Piz is a supercomputer of Switzerland. Supercomputer named Piz Daint derived from a mountain Piz Daint in Swiss Alps. Piz Daint is placed in Swiss National Supercomputing Center. Piz Daint in November 2016 was 3rd most efficient computer the world. Now according to ‘TOP500’ Piz ranked 5th in the world.
Design of Piz Daint:
In the very starting Cray XC30 was used in 2012. In October 2016 system increased with 1256 nodes with Piz Dora (Cray XC40). Now both Piz Daint & Piz Dora both are combined in a single unit Cray XC50 Nvidia Tesla P100 GPUs.
6. TRINITY (UNITED STATES)
IN 2013 NERSC (National Energy Research Scientific Computing Center) & ACES (Alliance for Computing at Extreme Scale) demands technical requirement of Trinity. And in November 2015 Trinity was at the top 6 in the list. But in November 2016 its rank fall to top 10 then Haswell and KNL (Knights Landing Partition) came together and Then Trinity again acquire its position on 6th place in November.
As above said Trinity was merged with KNL so there are 2 parts of its manufactures. In first part Intel Xenon Haswell processor and in Second part Xenon Phi Knights used to increase its performance. A 40 PFLOPS performance is given by total of 301952 Haswell and 678912 Knights Landing processors.
Design of Trinity:
In Trinity Cray XC40 is used with total memory capacity of 2.07 PiB. There are total 19420 computer nodes with a total 8.6MW power requirement. In Trinity there are total 5 initial storage groups:
Parallel File System
Physical memory of system is 2 PB DDR4 DRAM. Every single processor has its own DRAM that gives extra storage. Data in this stage is very short-term and reside only for minute time.
In this supercomputer there are 300 XC40 blades are data warp and each contain two burst buffer and four SSD drives. In this stage total storage is of 3.78 PB with transfer capability of 2 TB/s. In this stage data resist for some hours.
Parallel File System:
Sonexion based Lustre file systems used by trinity with total storage of 78. Transfer capability is nearly 1.8 TB/s in this stage. In this stage data is used for performing HPC activities. In this stage data resist for few weeks.
Campaign Storage stage is good for MarFS file system and also good for mix properties of POSIX and Storage models. This stage has a growing capacity at approximate rate of 30 PB/ annum. Present capacity of the system is more than 100 PB. with a current capacity of over 100 PB. During testing Scientist framed 968 billion files in one directory at a creation rate of 835 million/second.
Key Design goals
- Data protection
- Ease of administration
This is the last stage of storage. In this stage approximately 100 PB data can be stored.
7. AI BRIDGING CLOUD INFRASTRUCTURE (JAPAN)
This is the supercomputer of Japan with a capacity of 19.88 PFLOPS. ABC developed at the University of Tokyo. This supercomputer is used to perform many tasks like Machine learning, AI (Artificial Intelligence) & deep learning. This Supercomputer was developed by Japan’s National Institute of Advanced Industrial Science & Technology with a total power consumption of 3MW only.
Design of ABCI:
This supercomputer is designed for low precision floating point, high data & Artificial Intelligence not for LINPACK benchmark.
8. SUPERMUC-NG (GERMANY)
This is the supercomputer of Germany with a capacity of 19.477 PFLOPS. This supercomputer is built by Lenovo at Leibniz Supercomputer Center. Processor used in this supercomputer is Xenon Platinum 8174 with operating system on Linux (SLES). According to ‘TOP500’ SuperMUC ranked on 8th place in the list. This the only supercomputer of Germany which listed in best 10 supercomputer’s list.
9. TITAN (UNITED STATES)
Titan is the supercomputer of United states. This is the 4th supercomputer of US in the list of top 10. US have total of 5 supercomputers out of 10. But in the list of 500 China top the list. This Supercomputer was made by Cray at Oak Ridge National Laboratory for scientific purpose. Upgradation of Jaguar in 2011 resulted in Titan. With a capacity of 10 PFLOPS. Titan was fully upgraded in 2013 and was able to use. As stated above Titan is used for any scientific purpose mainly for Molecular Physics and climate Models. Titan was on the first place in November 2012 but in June 2013 Tianhe-2 took the place of Titan. In November 2018 Titan placed at 9th position in the list.
Design of Titan:
A total of 18688 CPUs are attached with the same number of GPUs to give the performance of 17.59 PFLOPS in LINPACK benchmark. This supercomputer is also termed as OLCF-3. Operating system of Titan is Linux (CLE, SLES)
10. SEQUOIA (UNITED STATES)
This is the 5th out of best 10 supercomputer of US. This supercomputer is built by IBM at Lawrence Livermore National Laboratory. IBM built this supercomputer for National Nuclear Security Administration. LINPACK benchmark for this supercomputer is 17.17 PFOLPS. Sequoia is an energy efficient computer; 7.9 MW energy is used by this computer. That’s why this supercomputer replace K supercomputer. In the starting Sequoia was used for Nuclear weapons but now it is used for various purposes like: Astronomy, Energy, lattice, study of human genome, and climate change. Latest use of Sequoia is in Fluid dynamics for solving complex problems of turbulence. In June 2013 it was behind Titan and Tianhe-2 But in November 2013 it dropdown to 4th place. Then in November 2016 it slipped again to 5th place and in November 2017 its position again fell down to sixth and in November 2018 its position came to last place in the list of top 10.
Design of Sequoia:
Sequoia design is based on Blue Gene Designs. 98304 computer nodes are used in this supercomputer. A2 processors chips with memory of 16GB each. So total memory of supercomputer is 1024*16 = 1572864 i.e. 1.5 PiB. Total area covered by this near about 280 square meters.