=Paper=
{{Paper
|id=Vol-2267/298-301-paper-56
|storemode=property
|title=Application of BOINC-based volunteer computing for comparison of the geoacoustic inversion algorithms efficiency
|pdfUrl=https://ceur-ws.org/Vol-2267/298-301-paper-56.pdf
|volume=Vol-2267
|authors=Oleg S. Zaikin,Pavel S. Petrov,Ilya I. Kurochkin
}}
==Application of BOINC-based volunteer computing for comparison of the geoacoustic inversion algorithms efficiency==
https://ceur-ws.org/Vol-2267/298-301-paper-56.pdf
Proceedings of the VIII International Conference "Distributed Computing and Grid-technologies in Science and
Education" (GRID 2018), Dubna, Moscow region, Russia, September 10 - 14, 2018
APPLICATION OF BOINC-BASED VOLUNTEER
COMPUTING FOR COMPARISON OF THE GEOACOUSTIC
INVERSION ALGORITHMS EFFICIENCY
O.S. Zaikin 1,2,a, P.S. Petrov 3,4, I.I. Kurochkin 5
1
ITMO University, 49 Kronverkskiy prospekt, Saint Petersburg, 197101, Russia
2
Matrosov Institute for System Dynsmics and Control Theory SB RAS, 134 Lermontov street, Irkutsk,
664033, Russia
3
V.I. Il’ichev Pacific Oceanological Institute FEB RAS, Far Eastern Federal University, 43
Baltiyskaya street, Vladivostok, 690041, Russia
4
Far Eastern Federal University, 8 Sukhanova street, Vladivostok,
690090, Russia
5
A.A. Kharkevich Institute for Information Transmission Problems RAS, 19 build 1 Bolshoy Karetny
pereulok, Moscow, 127051, Russia
E-mail: a zaikin.icc@gmail.com
The BOINC-based volunteer computing project Acoustics@home was employed to study the accuracy
of the sound speed profile reconstruction in a shallow-water waveguide using a dispersion-based
geoacoustic inversion scheme. This problem was transformed into a problem of black-box
minimization of a certain mismatch function. According to the first approach, a sound speed profile is
considered a piecewise-linear function with fixed uniformly-spaced nodes. At these nodes the values
of sound speed are obtained in the course of inversion. In the second approach the depths of the sound
speed profile nodes are also considered inversion parameters, however their number must be smaller
than in the first approach due to the computational complexity limitation. Several large-scale
computational experiments reveal that for the considered problem the second approach leads to a more
accurate sound speed profile estimation.
Keywords: volunteer computing, BOINC, geoacoustic inversion, underwater acoustics, black-box
optimization
© 2018 Oleg S. Zaikin, Pavel S. Petrov, Ilya I. Kurochkin
298
Proceedings of the VIII International Conference "Distributed Computing and Grid-technologies in Science and
Education" (GRID 2018), Dubna, Moscow region, Russia, September 10 - 14, 2018
1. Acoustics@home
Volunteer computing [1] is a type of distributed computing, that is based on the usage of
resources donated by private persons via the Internet. Volunteer computing suits well to solve
computationally hard problems, which can be decomposed into a set of independent subproblems. In
last two decades, several important results in various scientific areas (astronomy, medicine,
mathematics, etc.) were obtained in volunteer computing projects. The most popular platform for
organizing volunteer computing projects is BOINC (Berkeley Open Infrastructure for Network
Computing) [2].
In March 2017, the BOINC-based volunteer computing project Acoustics@home was
launched [3]. It is the first volunteer computing project aimed at solving computationally hard
problems in underwater acoustics. The project software is based on the MPI-program CAMBALA [4],
that is designed for launching on computing clusters.
Acoustics@home has several daemons, launched on the project’s server. Work generator
decomposes an original problem into independent subproblems by varying several considered
parameters. The rest parameters are varied in the computing application of the project which is
launched on volunteers’ computers.
Since its launch, the project has been used for solving hard inversion problems. In particular,
sound speed profile reconstruction in a shallow-water waveguide was performed. For this purpose, a
dispersion-based geoacoustic inversion scheme was employed. This scheme is briefly described in the
next section.
2. Geoacoustic inversion as a black-box optimization problem
The study of ocean bottom is required for mining, while the study of its water columns can be
used for underwater navigation and tracking of large sea animals. Using geoacoustic inversion, water
column and bottom parameters can be reconstructed from acoustic data [5]. Usually the data for the
geoacoustic inversion is mostly obtained using expensive receiver arrays.
Recently it was shown that a broadband pulse signal recorded by a cheap single hydrophone
can be also successfully used for estimating the acoustical parameters of sea bottom (see, e.g., [6]).
The inversion procedure in this case relies on the dependence of arrival times on frequency and mode
number (see details, e.g., in [6, 7]).
The implementation of this method in practice can be reduced to a black-box minimization
problem, in which an objective function should be minimized in a discrete search space [3, 4], and
every evaluation of this function requires numerous solutions of an acoustic spectral problem [4, 7].
Thus, the corresponding search space can be easily divided into a large number of relatively simple
independent tasks, which can be processed in parallel.
3. Computational experiments
In 2017, in Acoustics@home the problem of estimation of the sound speed profile in a water
column was considered. This problem is described in [3] in detail. A sound speed profile was
considered a piecewise-linear function with fixed uniformly-spaced nodes. The more nodes one can
afford, the better is the possible accuracy of the estimation based on the dispersion-based geoacoustic
inversion algorithm. In the first experiment, values of five nodes were varied. Each node had 31
possible sound speed values (from 1450 m/s to 1510 m/s with the step of 2 m/s), so the search space
contained 28 629 151 points. A brute-force algorithm was employed to calculate a value of the
objective function in each point of the search space. In each such point the corresponding direct
problem can be solved in few seconds on 1 CPU core. The search space was divided into 29 791
workunits, which were made by varying values of the first three nodes. Acoustics@home client
application had to vary values of other two nodes for each workunit, so 961 points had to be processed
within it. On average it took about 1 hour to process one such workunit on 1 CPU core. This
299
Proceedings of the VIII International Conference "Distributed Computing and Grid-technologies in Science and
Education" (GRID 2018), Dubna, Moscow region, Russia, September 10 - 14, 2018
experiment was completed within 10 days. The best value of the objective function was 0.00428371.
In the second experiment, values of six nodes were varied. The search space was divided into 923 521
workunits (by varying the values of the first four nodes), each of them consisted of 961 points. This
experiment took 58 days. The best value of the objective function was 0.00384831. In the considered
experiments the project’s performance was comparable to that of a computational cluster equipped
with 500 modern CPU cores.
In 2018 two more experiments were held to solve the same inversion problem. In these
experiments, the depths of the sound speed profile nodes were also considered inversion parameters.
In the third experiment, three sound speed profile nodes were considered. The depth of the first node is
constant (it corresponds to the water surface, i.e. it has the depth 0 m), while the depths of other nodes
were varied from 2 to 40 meters with the step 1 meter. In total, there were 666 possible combinations
of depths values. Values of all three nodes were varied too. While each of three nodes had 36 possible
values (from 1440 m/s to 1510 m/s with the step of 2 m/s), the search space contained 31 072 896
points (5 parameters were varied in total). 23 976 workunits were generated and processed within 12
days. As a result, the minimum of the objective function 0.003945 was found: The corresponding
point is: 23 m, 33 m, 1506 m/s, 1490 m/s, 1462 m/s.
In the fourth experiment, four nodes were considered. Depths of three nodes were varied with
the same bounds and step as in the third experiment. In total, 7 140 possible combinations of depths
values were constructed. While each node had 36 possible values, the search space contained
11 992 458 240 points (7 parameters were varied in total). The experiment was launched in April 2018
and is still running; 52 % of the search space has been processed within 6 months. The current
minimum of the objective function is 0.003784. The corresponding point is: 23 m, 34 m, 40 m, 1506
m/s, 1490 m/s, 1454 m/s, 1466 m/s. It should be noted, that this value of the objective function is the
best one for the considered problem found so far.
4. Acknowledgement
This study was partially supported by the Council for Grants of the President of the Russian
Federation (grant No. MK-2262.2017.5) and the Russian Foundation for Basic research (grants No.
16-07-00155-a, No. 18-29-03264-mk and No. 18-57-06003-Az-a). The authors thank all
Acoustics@home volunteers, whose computers took part in the experiments.
5. Conclusion
Several large-scale computational experiments were held in Acoustics@home. It was
revealed, that for the considered problem the approach, in which the depths of the sound speed profile
nodes are also considered inversion parameters, leads to a more accurate sound speed profile
estimation.
References
[1] Holohan A. Community, Competition and Citizen Science: Voluntary Distributed Computing
in a Globalized World. Global Connections. Taylor & Francis. 2016.
[2] Anderson D.P. BOINC: a system for public-resource computing and storage // Proceedings of the
5th IEEE/ACM International Workshop on Grid Computing, GRID’04. Washington, DC, USA, 2004.
IEEE Computer Society. pp. 4-10.
[3] Zaikin O.S., Petrov P.S, Posypkin M.A., Bulavintsev V.G., Kurochkin I.I. A volunteer computing
project for solving geoacoustic inversion problems // Open Engineering. 2017. Volume 7, Issue 1.
pp. 363-370.
300
Proceedings of the VIII International Conference "Distributed Computing and Grid-technologies in Science and
Education" (GRID 2018), Dubna, Moscow region, Russia, September 10 - 14, 2018
[4] Zaikin O.S., Petrov P.S. Algorithm of reconstruction of the sound speed profile in a shallow-water
geoacoustic waveguide from modal dispersion data // Optoelectronics Instrumentation and Data
Processing. Vol. 52, Issue 3. pp. 259-265.
[5] Jensen F.B., Kuperman W.A., Porter M.B., Schmidt H. Computational ocean acoustics. Springer,
New-York et al. 2011.
[6] Bonnel J., Chapman N.R. Geoacoustic inversion in a dispersive waveguide using warping
operators // The Journal of the Acoustical Society of America, 130(2):EL101–EL107. 2011.
[7] Petrov P.S. A method for single-hydrophone geoacoustic inversion based on the modal group
velocities estimation: Application to a waveguide with inhomogeneous bottom relief // Proceedings
of the International Conference Days on Diffraction. 2014. pp. 186–191.
301