Original papersDevelopment of an orchestration aid system for gridded crop growth simulations using Kubernetes
Introduction
Spatial assessment of climate change impact on crop production would aid identification of the adaptation measures suitable for a region of interest (Rosenzweig et al. 2014). It would be prudent to take into account both the outcome of local studies and crop growth simulations for the reasonable design of adaptation options (Beveridge et al. 2018). In particular, it would be preferable to use crop growth simulations over an extended area in order to overcome the limited number of field-based studies. For example, the gridded simulations of crop growth have been performed at global and regional scales (Asseng et al. 2019), which provides spatial information on the potential outcomes of adaptation options under future climate conditions.
A process-based crop model, which simulates the biophysical responses of crops to given conditions (He et al. 2017), has been used to evaluate the management practices chosen under a given climate condition. For example, Bassu et al. (2009) suggested that planting date shift would be a viable option to minimize the negative impact of climate change on crop production using crop growth simulations. Chun et al. (2016) illustrated that the benefit of planting date shift for rice production would differ by region through gridded crop growth simulations over Southeast Asia.
A large number of crop growth simulations would be needed to evaluate crop management options under environmental and management conditions (Jang et al. 2019). Stakeholders or local farmers would benefit from knowledge of planting dates and cultivars suitable for a given climate condition (Sacks et al. 2010). Still, the adaptation measures, e.g., planting date and cultivar, may differ by time and locality (Ishigooka et al. 2017). As a result, the number of crop growth simulations increases considerably to evaluate adaptation options over extended time periods and regions. This would require considerable computing resources and expertise for a large amount of the simulations. For example, Xiong et al. (2020) used a high performance computer to launch 1440 sets of the global gridded simulations with the combinations of climate projections, crop models, parameter strategies and sowing dates.
Various software has been developed to support the gridded simulations (Resop et al., 2016, Hyun et al., 2017, Yoo et al., 2018). Researchers can write their own scripts to deal with the tasks for the gridded simulations such as preparation of inputs, execution of model run and aggregation of outputs. Alternatively, integrated tools for the gridded simulations can be used to minimize efforts for such tasks (Shelia et al. 2019). However, these tools are often developed for a small set of gridded simulations, which supports limited functionality to minimize labor or accelerate computation for a large number of the gridded simulations.
Multiple sets of gridded crop growth simulations can be performed in parallel using the orchestration of workloads. The orchestration of the gridded simulations include configuration, coordination and management of containers, which enable customized environment and performance isolation (Rodriguez and Buyya 2019). For example, containers specialized for the gridded simulations can be deployed to a large number of computers using an orchestration system such as Kubernetes (Beltre et al. 2019). However, the demand for computer expertise could make it challenging for researchers to perform multiple sets of gridded crop growth simulations using the orchestration system.
The objective of our study was to develop a software framework that supports the orchestration of the gridded crop growth simulations. We also focused on implementation of the framework for researchers who have limited budget and expertise in the orchestration system. In particular, the orchestration aid system was applied to single board computers, which allowed for a considerably large number of crop growth simulations with minimum costs and efforts. Such a system would facilitate the spatial assessment of climate change impact on crop production and the design of adaptation strategies in a region.
Section snippets
Kubernetes
Kubernetes has been used to orchestrate multiple tasks using virtualized computing resources (Beltre et al. 2019). Kubernetes provides environments for management of a pod, which is a basic execution unit for a service using containers. For example, pods can be created within a physical computer to operate server or client tools for a distributed computing system. These pods remain active until all the gridded simulations are completed. When the service is no longer needed, these pods can be
Case study
A use case was set to perform multiple sets of gridded simulations for identification of an optimum planting date window for rice production in South Korea using GROWLERS-kube. The outcome of these simulations can be used to design an ideal cropping system at specific sites. For example, the most delayed planting date can be identified to escape an extreme weather condition, which could occur frequently in the future.
Wall time for the gridded simulations
The wall time for the gridded simulations was similar when the virtualized cluster computer sets had the same number of client pods. For example, V7S1C2 and V5SC2 completed the given sets of gridded simulations in a similar time although the former performed 40% more simulations than the latter (Fig. 6). The shared server pod did not delay the wall time for the given sets of the gridded simulations on average. For example, the D value for the Kolmogov-Smirnov test was 0.233 (p = 0.958) between
Discussion and conclusion
Our results demonstrated that the software framework for the orchestration of the gridded simulations can facilitate the spatial assessment of climate change impact on crop production under diverse crop management scenarios. GROWLERS-kube, which includes PodCreator and PodLauncher tools, allowed for a user to organize single board computers into a high performance computer without knowledge of an orchestration system such as Kubernetes. For example, GROWLERS-kube becomes operational for the
CRediT authorship contribution statement
Junhwan Kim: Conceptualization, Writing - original draft. Jin Yu Park: Software. Shinwoo Hyun: Visualization, Data curation. Byoung Hyun Yoo: Conceptualization, Data curation. David H. Fleisher: Writing - review & editing. Kwang Soo Kim: Funding acquisition, Methodology, Writing - original draft.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This study was supported by Cooperative Research for Agriculture Science Technology.
Development (PJ013837032019) program.
References (29)
- et al.
Optimising sowing date of durum wheat in a variable Mediterranean environment
Field Crops Res.
(2009) - et al.
Assessing rice productivity and adaptation strategies for Southeast Asia under climate change through multi-scale crop modeling
Agric. Syst.
(2016) - et al.
ASDF: A new data format for astronomy
Astron. Comput.
(2015) - et al.
Data requirement for effective calibration of process-based crop models
Agric. For. Meteorol.
(2017) - et al.
Development of an EPIC parallel computing framework to facilitate regional/global gridded crop modeling with multiple scenarios: a case study of the United States
Comput. Electron. Agric.
(2019) - et al.
The DSSAT cropping system model
Eur. J. Agron.
(2003) - et al.
Development of an automated gridded crop growth simulation support system for distributed computing with virtual machines
Comput. Electron. Agric.
(2020) - et al.
Development and application of a weather data service client for preparation of weather input files to a crop model
Comput. Electron. Agric.
(2015) - et al.
A multi-scale and multi-model gridded framework for forecasting crop production, risk analysis, and climate change impact studies
Environ. Modell. Software
(2019) - et al.
Development of a gridded climate data tool for the COordinated Regional climate Downscaling EXperiment data
Comput. Electron. Agric.
(2017)
Climate change impact and adaptation for wheat protein
Global Change Biol.
Crop modelling: towards locally relevant and climate-informed adaptation
Climatic Change
Performance comparison of lightweight kubernetes in edge devices, international symposium on pervasive systems
Algorithms Networks
Cited by (4)
A modularized parallel distributed High–Performance computing framework for simulating seasonal frost dynamics in Canadian croplands
2023, Computers and Electronics in AgricultureGLUEOS: A high performance computing system based on the orchestration of containers for the GLUE parameter calibration of a crop growth model
2022, Computers and Electronics in AgricultureCitation Excerpt :For example, Lamsal et al. (2018) used a super-computer to perform 356 million simulations for parameter calibration. However, the use of a high-performance computing system often requires a high level of expertise including computer programming and network management (Kim et al., 2021). In the present study, it was shown that the functionality of GLUEOS would allow for time-efficient parallel-processing based parameter calibration through a web interface, which requires no end-user expertise in distributed computing.
A Bibliometric Analysis of the Application of Remote Sensing in Crop Spatial Patterns: Current Status, Progress and Future Directions
2022, Sustainability (Switzerland)