![Page 1: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/1.jpg)
Efficient energy use of local and remote data processing
Team Name: MalaJunta
2018
Team Members: • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev • Nikita Volkov • Oleg Ziangirov
![Page 2: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/2.jpg)
Perfect Storm 2
Smartphone penetration is sky-rocketing
More devices are being created with Wi-Fi capabilities and sensors built into them.
The cost of connecting and technology are
going down
Broadband Internet is become more widely available.
Lower Power device IoT
![Page 3: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/3.jpg)
3
Network of physical devices, with electronics, software, sensors, and network connectivity which enable these objects to connect and exchange data.
IOT
![Page 4: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/4.jpg)
Where can we find it?It is everywhere
4
![Page 5: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/5.jpg)
The Problem 5
The full potentials of new wireless technology in the case of power saving has not been reached.
There are different reasons that affects the power consumption, one of theme is that the full data processing is not optimally distributed between the sensor and the host .
Lower Consumption Optimization
![Page 6: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/6.jpg)
Unbalance competition.
The jet engine fitted with 5,000 sensors can generate up to 10 GB of data per second.
Jet engine example
![Page 7: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/7.jpg)
What we will do: 7
We will implement a simple NLA data processing algorithm and compare the efficiency of it when:
We process the data in the host
We process the data in the smart
sensor
![Page 8: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/8.jpg)
Algorithm
Language: C 99 Library: arm_math.h
8
BLAS 1 A + B = C
data operations
BLAS 3matrix-by-matrix
A*B=C data
operations
BLAS 2matrix-by-vector
A*v = v’ data
operations
02 0301
Matrix-by-matrix product is the core for almost all efficient algorithms in linear algebra. (NLA Lecture 3, Ivan Oseledets)
![Page 9: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/9.jpg)
Methodology:Nrf52-dk + Power Profiler Kit nRF6707
9
![Page 10: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/10.jpg)
ResultsPower Consumption
10
Local Processor
Remote Processor
Mili
Am
per
3.1
![Page 11: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/11.jpg)
Conclusion & Future Steps 11
By processing the data in a local way we can save from 1 to 1.5 years of battery life time .
01
02
03
In base of the results we can select the most optimal way to process data according to the characteristics of the task and do it in an efficient way.
As future steps, we are hoping to find a criteria of effective data processing allocation.
![Page 12: Efficient energy use of local and remote data processingnla.skoltech.ru/projects/files/presentations/team_35.pdf · • Kupchenko Leonid • Jerry Gozie • Konstantin Chukreev •](https://reader035.vdocuments.us/reader035/viewer/2022081611/5f06e7f47e708231d41a52b3/html5/thumbnails/12.jpg)
Sensor to use:CC2650 Wireless MCU
Contains sensors for:
12
Gyroscope
Temperature
Digital Microphone
Light
Accelerometer
Pressure
Humidity
Magnetometer