Slides used for presentation at PEARC17 of the conference paper of the same name

1. CyberGIS Center for Advanced Digital and Spatial Studies
A CyberGIS-Jupyter Framework for
Geospatial Analytics at Scale
Dandong Yin1,2, Yan Liu1,2,3, Anand Padmanabhan1,2,3,
Jeff Terstriep1,3, Johnathan Rush1,3, and Shaowen Wang1,2,3
1CyberGIS Center for Advanced Digital and Spatial Studies
2Department of Geography and Geographic Information Science
3National Center for Supercomputing Applications (NCSA)
University of Illinois at Urbana-Champaign
Practice & Experience in Advanced Research Computing (PEARC17)
July 11, 2017, New Orleans, Louisiana, USA

2. CyberGIS Center for Advanced Digital and Spatial Studies
Background
• Gateways facilitate and accelerate scientific research and education enabled by
advanced cyberinfrastructure (CI)
• End-user-oriented development approaches often consume significant resource
and time to achieve these desirable goals
2
Advanced Cyberinfrastructure
Gateway Gateway Gateway
• To fully leverage CI, it is necessary to not
only reduce the barrier of accessing CI via
gateways, but also reduce the barrier of
developing gateway applications so that
common researchers could efficiently
deliver their own applications to their
communities

3. CyberGIS Center for Advanced Digital and Spatial Studies
End-user-orientedmodel
• Final products need to be of fine quality in order to serve diverse users
3Topolens: A CyberGIS Gateway Application
• Usually implemented with
web-application frameworks
• Frontend UI design
• Backend computation
• Database design
• Costly to develop, optimize,
and maintain
• Limited to specific purposes
• Difficult to extend and re-use

4. CyberGIS Center for Advanced Digital and Spatial Studies
A new gateway paradigm
• In many scientific research scenarios, agility, reproducibility and extensibility are
increasingly important
4
Advanced Cyberinfrastructure
• Community-driven gateway development
• A series of building blocks to be re-organized for
new applications
• A public platform for domain scientists to
develop, reproduce and share their own
analytics at scale
Gateway Gateway
Gateway
• Common and friendly user interfaces for
collaborators, researchers, and gateway
developers

5. CyberGIS Center for Advanced Digital and Spatial Studies
CyberGIS
• CyberGIS -- geographic information
science and systems (GIS) based on
advanced CI
• Innovate new-generation GIS
• Focus on computational and data-
intensive geospatial problem-solving
within various research and education
domains
• Bridge gaps between geospatial big data,
software and applications through
• innovative cyberGIS supercomputer --
ROGER
5
Wang, S. (2010) A CyberGIS Framework for the Synthesis of
Cyberinfrastructure, GIS, and Spatial Analysis. Annals of the Association
of American Geographers, 100(3): 535-557

6. CyberGIS Center for Advanced Digital and Spatial Studies
• An interactive scripting interface
• Declarative UI widgets
• Supports general computation and
visualization
• Increasingly popular in data- and
computational sciences
Jupyter Notebook
6

7. CyberGIS Center for Advanced Digital and Spatial Studies
• An interactive scripting interface
• Declarative UI widgets
• Supports general computation and
visualization
• Increasingly popular in data- and
computational sciences
CyberGIS-Jupyter
7

8. CyberGIS Center for Advanced Digital and Spatial Studies
CyberGIS-Jupyter: Technologies
8
Batch Job management
Cloud-based architecture
Data/storage synchronization

9. CyberGIS Center for Advanced Digital and Spatial Studies
CyberGIS-Jupyter: Innovation
• Geovisualization
• Interactive map generation inside notebooks
• Support multiple formats of geospatial data
• Layer management, transparency and styles
9
No existing technologies found that match all the
criteria above. Therefore, we developed a library
named Floret to serve the purpose (shown in
demo).

10. CyberGIS Center for Advanced Digital and Spatial Studies
Features
• Adaptive interfaces for CI access embedded in notebooks
• Programming or GUI at choice
• Manage batch jobs at fingertips
• Cloud-based environment
• Reproducible computing environments with Docker
• Elastic resource-provisioning with OpenStack
• Shared storage
• Big data support
• Fault tolerance
• Collaborative sharing
• Generative geovisualization by programming
• Embedded web-mapping services
• Support multiple geographic data formats
10

11. CyberGIS Center for Advanced Digital and Spatial Studies
Architecture - 1
• Gateway applications are standardized as packages deployed on hybrid HPC
11

12. CyberGIS Center for Advanced Digital and Spatial Studies
Architecture - 2
12

13. CyberGIS Center for Advanced Digital and Spatial Studies
• Highly extensible
• Predefined UI templates
User interface
13

14. CyberGIS Center for Advanced Digital and Spatial Studies
• PAM with LDAP + SSSD
• Whitelist of authorized users
Authentication and authorization
14

15. CyberGIS Center for Advanced Digital and Spatial Studies
• Using JupyterHub with swarm spawner
• Openstack VMs join/leave swarm as needed
Container management
15

16. CyberGIS Center for Advanced Digital and Spatial Studies
• Widgets for composing job scripts (PBS files)
• Submit and retrieve information (via qsub and qstat)
Computation management
16

17. CyberGIS Center for Advanced Digital and Spatial Studies
• GPFS -> NFS -> VM Volume -> Container Volume
• Enable volume expansion, fault-tolerance and low-latency feedback
Data persistency
17

18. CyberGIS Center for Advanced Digital and Spatial Studies
Case study
18
• Height Above Nearest Drainage (HAND) at 10m for continental US
• Flood analysis map derived from 10m USGS 3DEP national elevation dataset (180 billion cells) and
National Hydrography Dataset (2.67 million stream reaches)
• Goals
a. Collaborative methodology development
b. HPC for data analytics
c. Deliver methodology and data products to
different user communities
• Collaborators
• Researchers
• Decision makers
• Students

19. CyberGIS Center for Advanced Digital and Spatial Studies
• Input data hosts
• ROGER GPFS storage
• Docker image
• GDAL, GEOS, PROJ4,
• HDF4, SQLite, MPI, etc.
• Sample notebook
• Explicit computation steps
• Seamless scaling up
• Output visualization
• Floret maps
HAND application package
19

20. CyberGIS Center for Advanced Digital and Spatial Studies
Comparison - 1
Collaboration
Requirements
Traditional Enhanced
Methodology
development
● Writeups as shared documents
● Communicated via emails, teleconf
● Methodology notebooks
○ Math formula
○ Code snippets
○ Sample results
Software
development
● Source codes only; computing
environment needs to be maintained and
synchronized manually
● Function notebooks with both source
codes and computing environments
effectively synchronized between
collaborators
Computation
● Conducted by a dedicated person, a
bottleneck
● Notebook interface to workflow
computation on advanced CI
○ Everyone can launch
20

21. CyberGIS Center for Advanced Digital and Spatial Studies
Comparison - 2
Result validation
● Data: direct download or via iRODS
● Validation results: shared document in
Google Drive
● Integrated validation notebooks with
reproducible input, statistics, and output
Visualization
● Local: download and use desktop GIS.
Not scalable for large outputs.
● Online: Tile Map Service (TMS); web-
GIS. Only available for major output data
● Integrated data, code, and visualization
notebooks
○ Traditional visualization libraries
○ Jupyter IFrame cells
○ CyberGIS Floret library
Collaboration
Requirements
Traditional Enhanced
21

22. CyberGIS Center for Advanced Digital and Spatial Studies
Feedbacks
• The CyberGIS-Jupyter framework was a major platform that supported UCGIS
summer school 2017
• Received impressively positive feedbacks from summer school attendees and
organizers
• According to an on-site survey, over 80% attendees strongly confirm the
importance of CyberGIS-Jupyter to enhancing cyberGIS-enabled research and
education
22

23. CyberGIS Center for Advanced Digital and Spatial Studies
Conclusions
• A new gateway development framework for facilitating and accelerating CI-
enabled research and education
•Domain scientists can easily adapt and create their own applications
•Students/collaborators can easily understand, reproduce and extend such gateway capabilities
• Leveraging cutting-edge technologies
•Jupyter notebooks and associated interactive interfaces
•Cloud-based elastic architecture
• To boost geospatial analytics and discoveries at desirable computing and
community scale
23

24. CyberGIS Center for Advanced Digital and Spatial Studies
Acknowledgements
• This work is supported in part by the National Science Foundation (NSF) under
grant numbers 1047916 and 1443080. The computational work used the NSF-
supported ROGER supercomputer (1429699).
• This work is also supported in part by the ECSS program of XSEDE, which is
supported by NSF grant number 1053575.
24

25. CyberGIS Center for Advanced Digital and Spatial Studies
DEMO
https://jupyter.cybergis.tk
25