1. Economical Supercomputing thru
Smartphone Crowd Computing
An Assessment of Opportunities, Benefits, Deterrents,
and Applications from India’s Perspective
Pijush Kanti Dutta Pramanik*
, Prasenjit Choudhury*
, Anindita Saha#
*
Dept. of Computer Applications, National Institute of Technology, Durgapur, West Bengal
#
Dept. of Computer Sc. & Engineering, B.T. Kumaon Institute of Technology, Dwarahat, Uttarakhand
2. POINTS TO BE DISCUSSED
• INTRODUCTION
• HPC POTENTIAL OF CONTEMPORARY SMARTPHONES
• SMARTPHONE MARKET AND USER DEVELOPMENT IN INDIA
• SUPERCOMPUTING USING SMARTPHONES
• BENEFITS
• DETERRENTS AND CHALLENGES
• SCOPE FOR HPC AND SUPERCOMPUTING IN INDIA
• OTHER SPECIFIC APPLICATIONS IN INDIAN PERSPECTIVE
• CONCLUSION AND FUTURE WORK
3. INTRODUCTION
• Buying and maintaining supercomputers are
exceedingly costly affairs
• Govt. of India has set to develop 70 supercomputers
with average cost of 60 crores per system with₹
additional 1,000 crores for power expense₹
• Desktop Grid Computing has been considered the
best alternative but uses of desktops are decreasing
rapidly
• Smartphones can give potentially equivalent
performance to a supercomputer
4. HPC POTENTIAL OF CONTEMPORARY
SMARTPHONES
• Deep Blue
– supercomputer that defeated Gary Kasparov
– delivered 11.38 GFLOPS
– whereas Samsung Galaxy S7 loaded with Adreno 530 GPU
offers 588 GFLOPS
• Tegra X1
– the first Tera-FLOPS SoC from NVIDIA
– more powerful than the ASCI Red, the first Tera-FLOPS
supercomputer of 2000, employed by the U.S. Department
of Energy’s Sandia National Laboratory
5. HPC POTENTIAL OF CONTEMPORARY
SMARTPHONES
• The enablers
– Symmetrical Multiprocessing (SMP)
– Heterogeneous Multi-Processing (HMP)
– GPU Accelerated Computing
– Cost-effective and less power-consuming SoCs (System on Chip)
– 64 bit octa-core CPUs (e.g. Snapdragon 820, Exynos 8890, Apple
A9X, HiSilicon’s Kirin 955) having clock frequencies not less than
1.5 GHz
– GPUs (e.g. GT7900, Adreno 530, ARM Mali-G71) having clock
frequencies greater than 500 MHz and delivers in the range of
300 – 800 GFLOPs
6. SMARTPHONE MARKET AND USER
DEVELOPMENT IN INDIA
• In 2011 more smartphones were shipped than PCs
• India and China are leading in smartphone uses
• By 2019, 51.5 percent (roughly 1.5 billion) of the
mobile population will be in this region
• In coming years, smartphone usage is expected to be
highest in India, among all Asia-Pacific countries.
• India has already suppressed the USA by crossing 220
million active smartphone users and has become
second highest only behind China
8. SUPERCOMPUTING USING
SMARTPHONES
• People use their smartphones generally for very few
hours in a day
• Huge amount of computing cycles are wasted
• Smartphone Crowd Computing (SCC)
– The computing power of potentially equivalent to a
modern supercomputer can be attained by properly
utilizing collective processing powers of these potent
smartphones
– An ad-hoc grid of available smartphones are formed
(opportunistically)
– Needy clients have to hook on to the grid to avail required
resource, not bothering about the intricacy of resource
accessing
9. SUPERCOMPUTING USING
SMARTPHONES
To implement SCC effectively, a distributed computing
framework is required to pull off the jobs like:
•Discovering suitable participant smartphone
•Distributing and scheduling jobs to the designated smart-
phones
•Collecting the results from multiple smartphones, updating in
the server application and assemble them for further purposes
•Implementing failure recovery schemes in case of failures.
10. BENEFITS OF SMARTPHONE CROWD
COMPUTING
• Economical
– No upfront investment (crowd sourced)
– No or minimal operational and maintenance cost
(including power and cooling cost)
• Environment-friendly computation
– ASCI Red gulped 500,000 watts of power with another
500,000 watts just to cool the building it sat in whereas
the Tegra X1 sips less than 15 watts of power altogether
– Smartphone processors are inherently energy efficient
• Flexible
– Anywhere an ad-hoc HPC can be setup
11. DETERRENTS AND CHALLENGES
• Limited battery power
• Heat
• Limited memory
• Connectivity and bandwidth
• Privacy and security
• Reliability
• People’s participation
12. SCOPE FOR HPC AND SUPERCOMPUTING
IN INDIA
• Biotechnology
• Atmospheric and oceanic sciences
• Astronomy and astrophysics
• Nuclear physics
• Space research
• Oil and natural gas exploration
• Meteorology
• Geology
• Mathematical modeling and simulation
• Weather forecasting
• Aeronautical development
• Manufacturing sector, especially automotive (for design and testing)
• Finance market (banks and stock exchanges)
• Big Data processing and analytics
13. OTHER SPECIFIC AREAS WHERE
SMARTPHONE CROWD COMPUTING
CAN BE UTILIZED
• Educational and research institutes
– Carrying out computing intensive researches
• Trains (and buses)
– Implementing collision detection avoidance methods (e.g. RCAS,
TCAS) in Indian railways
• Airways
– In-flight data processing
• Soldier’s outpost at difficult terrains
– Processing and analyzing data from sensor based gizmos
• Shopping mall
– Real-time product recommendation
– ERP and Mall management
14. CONCLUSION AND FUTURE WORK
• Smartphone crowd computing has great HPC
potential, especially in those situations where
traditional HPC setup is not feasible or economical
• Proper architecture and operating environment need
to be defined and framed
• Issues and challenges are to be considered more
thoroughly and application specific
– Resource discovery
– Job scheduling
– Job offloading
– Defining service level agreement
15. Citation and conference paper link:
Pijush Kanti Dutta Pramanik, Prasenjit Choudhury
and Anindita Saha, “Economical Supercomputing
thru Smartphone Crowd Computing: An Assessment
of Opportunities, Benefits, Deterrents, and
Applications from India’s Perspective”, 4th
International Conference on Advanced Computing
and Communication Systems (ICACCS - 2017), January
06-07, 2017, Coimbatore, India, IEEE
[DOI: 10.1109/ICACCS.2017.8014613]
https://ieeexplore.ieee.org/document/8014613/
Editor's Notes
Desktop Grid Computing has been considered as an inexpensive alternative of the highly-priced supercomputers. The popularity of desktops as computing devices is fading quickly in the age of powerful smartphones.
Today’s smartphones can deliver equivalent or even better performance than supercomputers of yesteryears.
The computing power of potentially equivalent to a modern supercomputer can be attained by properly utilizing collective processing powers of these potent smartphones.
SMP – each core share workloads, run at a lower frequency which in turn results in less power consumption
HMP - CPUs comprise of two different sets of cores. The aim is to use ‘LITTLE’ cores for most of the time and use the ‘big’ cores only for high-frequency operations. significant average power savings without compromising the peak performance.
GPU Accelerated Computing - compute-intensive segment of the application is passed on to the GPU; however, rest of the code is executed on the CPU. GPU-accelerated computing can deliver exceptional performance if the right process is scheduled to the right core.
SoC has made it materialize to put a whole computer on a chip and reduced the size to a thumbnail.
SMP – each core share workloads, run at a lower frequency which in turn results in less power consumption
HMP - CPUs comprise of two different sets of cores. The aim is to use ‘LITTLE’ cores for most of the time and use the ‘big’ cores only for high-frequency operations. significant average power savings without compromising the peak performance.
GPU Accelerated Computing - compute-intensive segment of the application is passed on to the GPU, rest of the code is executed on the CPU. GPU-accelerated computing can deliver exceptional performance if the right process is scheduled to the right core.
SoC has enabled to put a whole computer on a chip and reduced the size to a thumbnail.
1. The Asia-Pacific region, especially China and India will have the maximum number of smartphone users in coming years.
2. eMarketer, a global marketing research and intelligence firm, estimates that by 2019, 51.5 percent (roughly 1.5 billion) of the mobile population in this region will have smartphones [24].
3. In coming years, smartphone usage is expected to be highest in India, among all Asia-Pacific countries.
4. The primary reason behind the explode in smartphone market in India is the penetration of indigenous low-cost smartphone makers like Micromax, Carbon, Intex etc. along with cheap but potent Chinese products.
1. The Asia-Pacific region, especially China and India will have the maximum number of smartphone users in coming years.
2. eMarketer, a global marketing research and intelligence firm, estimates that by 2019, 51.5 percent (roughly 1.5 billion) of the mobile population in this region will have smartphones [24].
3. In coming years, smartphone usage is expected to be highest in India, among all Asia-Pacific countries.
4. The primary reason behind the explode in smartphone market in India is the penetration of indigenous low-cost smartphone makers like Micromax, Carbon, Intex etc. along with cheap but potent Chinese products.
1. People use their smartphones generally for very few hours in a day [28][29], which accounts a large number of unused smartphones at a certain point of time.
2. If, say, 60 or more smartphones with GFLOPS higher than 500 (e.g. Imagination PowerVRGT7900 or NVIDIA Tegra) are connected together, the combined GFLOPS can be well equivalent to some of the India’s top 30 supercomputers [8],minus the huge cost. If the number of such smartphones is increased to 100 or more, the accumulated performance can challenge some of the mighty supercomputers in the world [30].
1. Desktop Grid Computing offers a perfect cheap alternative.
2. The popularity of desktops is decreasing; instead, smartphones are becoming primary computing device to people thanks to extensive advancement in their computing ability.
3. Smartphone Crowd Computing (SCC) can aptly take the place of Desktop Grid Computing as the inexpensive alternative of costly supercomputers.