Cloud Revolution: Exploring the New Wave of Serverless Spatial Data
0 likes442 views
The document discusses the cloud revolution in geospatial data management and introduces cloud-native formats for spatial data, emphasizing efficiency and accessibility. It highlights various cloud-optimized formats, like COG and Zarr, which support dynamic and partial reads to enhance data processing and user experience. The content is framed around the challenges, benefits, and future opportunities of implementing these technologies within cloud-native spatial data infrastructures.
Cloud Revolution: Exploring the New Wave of Serverless Spatial Data
- 1. Cloud Revolution: Exploring the New Wave of Serverless Spatial Data
- 2. From Zip, Clip and Ship… To Read WhatYou Need! ✂🤐🛳
- 3. Dean Hintz Technical Support Team Lead, Strategic Solutions Safe Software Kailin Opaleychuk Technical Support Specialist, FME Desktop Safe Software Dale Lutz Co-Founder Safe Software
- 4. Michelle Roby Developer Advocate Radiant Earth Chris Holmes VP of Product, Strategy, Partnerships Planet
- 5. Agenda 1 Introduction 2 What is Cloud Native? 3 Perspectives from Radiant Earth, Planet 4 STAC & COGs 5 FlatGeoBuf 6 COPC & Zarr 7 Bonus… GeoParquet 8 Lessons learned 9 Q&A Agenda
- 6. Welcome to Livestorm. A few ways to engage with us during the webinar: Audio issues? Click this for 4 simple troubleshooting steps.
- 8. Cape Town, South Africa • March 19, 2017 Planet / Cloud Native Geo Foundation / Taylor Geospatial Engine Cloud Native Geospatial Origins Chris Holmes
- 35. Cloud-Optimized Data Formats Format Data Type Cloud-Optimized GeoTIFF (COG) Raster Zarr, Kerchunk Multi-dimensional Raster Cloud-Optimized Point Cloud (COPC), Entwine Point Tiles (EPT) Point Clouds* FlatGeobuf (FGB), GeoParquet (GPQ) Vector *Vector formats can do point clouds (spatial index for s varies). The line between needing a point cloud-specific format vs a vector format is blurry.
- 48. AGAVE PLANTATIONS • Tequila, Mexico • November 22, 2021 Towards Cloud-Native Spatial Data Infrastructure
- 50. AIRPORT • Shuttleworth, Birmingham • April 9, 2020 An SDI is a coordinated series of agreements on technology standards, institutional arrangements, and policies that enable the discovery and use of geospatial information by users and for purposes other than those it was created for. - Kuhn (2005)
- 52. © 2023 PLANET LABS PBC ALL RIGHTS RESERVED Benefits of CNSDI approach Scale Ease Cost
- 53. © 2023 PLANET LABS PBC ALL RIGHTS RESERVED
- 55. $29.47
- 58. Thank you!
- 60. Thank you Chris Holmes VP of Product, Strategy, Partnerships Planet
- 62. About Radiant Earth About: ● An incubator of data-driven initiatives, services, and 21st century institutions needed to foster shared understanding of our world Initiatives: ● Cloud-Native Geospatial Foundation → Aim to increase adoption of highly efficient approaches to working with geospatial data on the Internet. ● Source Cooperative → Data publishing utility for easy data sharing over the web. Introduction to Cloud-Optimized Formats
- 63. What does “cloud-optimized” mean? File formats are read-oriented to support: ● Partial reads ● Parallel reads ● (File) metadata in one read Cloud implementation also includes: ● Accessible over HTTP using range requests ● Supports lazy access and intelligent subsetting ● Integrates with high-level analysis libraries and distributed frameworks
- 64. Providers ● Less Downloads ○ Reduced Cost ○ Reduced Server Load ○ Sometimes smaller storage (If it’s a compressible format) ○ Serve more people with the same resources ○ Colocate Compute with Data Users ● Less Downloading ○ Less time waiting for data ○ Less time tossing out irrelevant data (masking) ○ Less data to load into memory ○ Less downloaded files to manage ■ Less storage ■ Less files ○ Bring the compute to the data Opportunities ● Serverless, Dynamic Tiling, Cloud Computing, New future awesome stuff! Why cloud-optimize your data?
- 65. Pre-rendered overviews or indexes Read-oriented file structure COG FGB Read what you need What is cloud-optimized data?
- 66. What makes cloud-optimized challenging? ● Many existing geospatial data storage formats ○ While all Earth observation data is “remotely sensed”, this data may be processed into raster, vector, and point cloud data types and stored in a long list of data formats and structures. ● User-dependent ○ Users must learn new tools and which data is accessed and how may differ depending on the user.
- 67. What makes cloud-optimized challenging? From Task 51 Study: “There is no one-size-fits-all packaging for data, as the optimal packaging is highly use-case dependent.” Authors: Chris Durbin, Patrick Quinn, Dana Shum
- 68. Poll: Which Cloud Native formats are you currently using?
- 69. Cloud-Optimized Data Formats Format Data Type Cloud-Optimized GeoTIFF (COG) Raster Zarr, Kerchunk Multi-dimensional Raster Cloud-Optimized Point Cloud (COPC), Entwine Point Tiles (EPT) Point Clouds* FlatGeobuf (FGB), GeoParquet (GPQ) Vector *Vector formats can do point clouds (spatial index for s varies). The line between needing a point cloud-specific format vs a vector format is blurry.
- 70. Cloud-Optimized Data Formats Format Data Type Replaces Adoption Standard Status Cloud-Optimized GeoTIFF (COG) Raster GeoTIFF Widely adopted (GDAL 3.1 Supported*) OGC* standard (October of this year) Zarr, Kerchunk Multi- dimensional Raster HDF5/netcdf 4 Adopted in particular communities (ie. Climate Science) (GDAL 3.4 Supported*) OGC standards in development Cloud-Optimized Point Cloud (COPC), Entwine Point Tiles (EPT) Point Clouds* las/laz Increasingly common (PDAL 2.4 Supported*, Entwine) 1.0 Specification FlatGeobuf (FGB), GeoParquet (GPQ) Vector shp/gpkg /geojson Increasingly common, Relatively new 🔥🔥🔥 (OGR 3.1, 3.5 Supported) OGC standards in development *OGC: Open Geospatial Consortium
- 71. ● COGs are raster data representing a snapshot in time of gridded data, for example digital elevation models (DEMs). ● The standard specifies conformance to how the GeoTIFF is formatted, with additional requirements of tiling and overviews. band Raster: COG (Cloud-Optimized GeoTIFF)
- 72. ● COGs have internal file directories (IFDs) which are used to tell clients where to find different overview levels and data within the file. ● Clients can use this metadata to read only the data they need to visualize or calculate. ● This internal organization is friendly for consumption by clients issuing HTTP GET range request ("bytes: start_offset-end_offset" HTTP header) Raster: COG (Cloud Optimized GeoTIFF)
- 73. ● Zarr is used to represent multidimensional raster data or “data cubes”. For example, weather data and climate models. ● Chunked, compressed, N-dimensional arrays. ● The metadata is stored external to the data files themselves. The data itself is often reorganized and compressed into many files which can be accessed according to which chunks the user is interested in. band Multi-dimensional Raster: Zarr
- 74. Multi-dimensional Raster: Kerchunk ● Kerchunk is a way to create Zarr metadata for archival formats, so that you can leverage the benefits of partial and parallel reads for archives in NetCDF4, HDF5, GRIB2, TIFF and FITS. ● Kerchunk negates the need to create and store copies of data for cloud-optimized access.
- 75. ● Columnar data, follows Parquet standards. ● 2 additions of GeoParquet to Parquet: encode geometries & include metadata. ● Highly compressed. ● Single-file or multi-file. ● No spatial-indexing (yet!). Vector: GeoParquet
- 76. COPC (Cloud-Optimized Point Clouds) ● Point clouds are a set of 3-dimensional (x,y,z) data points in space, such as gathered from LiDAR measurements. ● COPC is a valid LAZ file. ● Similar to COGs but for point clouds: COPC is just one file, but data is reorganized into a clustered octree instead of regularly gridded overviews. ● 2 key features: ○ Support for partial decompression via storage of data in a series of chunks ○ Variable-length records (VLRs) can store application-specific metadata of any kind. VLRs describe the octree structure. ● Limitation: Not all attribute types are compatible.
- 78. CNG Foundation Activities Include: ● “Holding the space” ● Development sprints ○ Held STAC Sprint #8 in Sep ○ Upcoming: ■ GeoParquet Community Day in San Francisco: Jan 30 ■ Zarr Sprint in NYC: Feb 7 - 8 ● Paid fellowships for software developers ○ Brandon Liu (@bdon) ● Sponsored feature development ○ HTTP extension for Zarr ● Documentation, tutorials, and other educational content (including webinars) ○ https://guide.cloudnativegeo.org ○ STAC webinar with Kenya Space Agency ○ CNG international webinar series (Brazil, Pacific Region, Africa)
- 80. Thank you Michelle Roby Developer Advocate Radiant Earth
- 81. “allow users to stream just the portion of data that’s needed, improving processing times and creating workflow opportunities that were previously not possible” ● Lower the bar for publishing your data ● Target key emerging cloud native formats ● Cover a range of data types: from imagery & point clouds to time series & vector data ● Streamline support across hybrid environments ● Leverage built in optimizations such as reader side filtering, feature tables, lazy evaluation Support FME users with easy access to data wherever it may be Safe’s Cloud Native Strategy
- 82. New Format Support Format Support Version Available Cloud Optimized Geotiff R / W 2023.0 Cloud Optimized Point Cloud R / W 2023.1 / 2023.2 FlatGeoBuf R / W 2023.0 GeoParquet R / W 2023.1 SpatioTemporal Asset Catalog (Metadata + Asset) R 2024.0 (FME Hub)* ZARR R / W 2023.1
- 84. STAC Package (FME Hub) - STAC Package V2.0.0 now available on the FME Hub. - STAC Metadata Reader* - STAC Asset Reader - V2.0.0 requires FME 24.0 minimum build 24134
- 85. STAC Metadata Reader - Images demonstrating how to use the STAC Metadata Reader to dig down into a STAC Collection https://spot-canada-ortho.s3.amazonaws.com/catalog.json
- 86. Slide Title Consume a GeoTIFF in STAC and convert to Cloud Optimized GeoTIFF Goal Key Result Working with STAC Asset Reader in FME Form Use the FME platform to refine and translate data from one location to another Output Cloud Optimized Geotiff ready for further analysis on S3
- 87. Demo
- 89. ● Use raster transformers to post-process STAC assets ○ Combining raster bands ○ Setting & removing no data ● FME’s S3Connector can publish COGs to the cloud Summary Removing no data FME Form Workspace
- 90. Demo Results
- 92. COG Reader - Search Envelope - Pyramid level options COG Writer - Writer feature type - Compression - Layout: Cloud Optimized Tiles - Pyramid level options
- 93. COG Reader in FME Form https://sentinel-cogs.s3.us-west-2.amazonaws.com/sentinel-s2-l2a-cogs/36/Q/WD/2020/7/S2A_36QWD_20200701_0_L2A/TCI.tif
- 94. COG Reader - Search Envelope Reading entire dataset Reading with Search Envelope constraint
- 95. Slide Title Create an insightful report on recent fires West of Kelowna Goal Key Result Current Fire Mapping for West Kelowna Use transformers to extract, combine & reformat data An interactive HTML report with embedded images and links
- 96. Demo
- 98. ● FlatGeoBuf and COG readers support spatial filter operations ● Use polygon mask to refine points on Nodata areas ● XMLTemplater can be used to help format HTML tables Lesson Slide
- 99. Demo Results
- 100. 4 FlatGeoBuf
- 101. FlatGeoBuf Reader ● Verify file buffers ● Search envelope FlatGeoBuf Writer ● Create spatial index
- 102. Slide Title Create a service that automatically uploads a range of vector data to S3 as FlatGeoBuf Goal Key Result FlatGeoBuf S3 Uploader App Generic Reader paired with user parameters Uploaded buffers and an upload html upload report
- 103. Demo
- 104. ● User parameters help make workspaces more dynamic ● PROJReprojector with online grids enabled Summary of FlatGeoBuf S3 Uploader App
- 106. ● Point cloud storage optimized for the web ● Only read what you need. This is especially powerful for point clouds given 3d data data volumes can be huge ● Query XYZ min/max ● Built on the LAS standard. ● Essentially uses the LAS reader / writer but with the COPC structure COPC
- 107. 6 ZARR Format
- 108. ● Multidimensional raster array / time series storage optimized for the web ● Based on NetCDF / HDF data cube formats ● Only read what you need ● Particularly powerful for raster time series, as multidimensional arrays often mean huge volumes ● Query XYT extents ● Zarr reads cube with each time step as a separate band with properties - easy to work with ZARR
- 109. ● Time series raster storage optimized for the web ● Based on NetCDF data cube ● NetCDF reads cube as multigrid with 1 band for each time step (hundreds of bands) and properties in attribute lists ● Zarr reads cube with each time step as a separate band with properties - easier to work with ● Default translation from NetCDF to Zarr just works* NetCDF to ZARR
- 110. ZARR CMIP5 Climate Model Temp Analysis: Winnipeg, MB
- 111. ZARR CMIP5 Climate Model Temp Analysis: Winnipeg, MB
- 112. OGC Climate Resilience Pilot 2023 Pilot Goals: ● Build climate resilience ● Expand audience for climate services ● Demonstrate the value of OGC standards and SDI’s (FAIR) ● Show how OGC can support international climate change goals ● Build a community of stakeholders better understand the range of possible impacts - allows us to better prepare and compensate for them https://www.ogc.org/initiatives/crp/
- 113. How to provide the data needed for climate impact and disaster indicators to a wider audience? ● Goal: Connect Climate and Disaster Pilots ● Data: Current situational awareness ○ Base map: physical, land use, infrastructure, pop ○ EO data: hazards and impacts ○ Drought & hydrologic monitoring ● Data: Future change awareness - risk scenarios due to climate change ○ Climate model outputs - time series data cubes ○ Temperature, precipitation and moisture projections ○ Analysis Ready Data (ARD) model results summary ○ Climate services known in climate community but not well known or utilized across affected impact domains NetCDF from Environment Canada Disaster Pilot 2023: Disaster and Climate Data Sources to ARD & Impacts
- 114. MB Drought Risk: Combined Precip Temp Query OGC API Features Query Parameters: Start Year: 2020 End Year: 2060 BBox: -100.0,49.0,-96.0,50.5 Limit: 2,000,000 MinPeriodValue: 0 (PrecipDelta) MaxPeriodValue: 0.75 (PrecipDelta) MinTemp: 23C (Min Mean Monthly Temp) Find all time step points over the next 40 years for southern Manitoba where projections indicate: ● > 25% dryer than historical mean AND ● mean monthly temperature > 23C
- 115. MB Precipitation: Future Delta PrecipDelta = PrecipFuture / PrecipHistoricalMean / = Yields normalized value from 0 to N where 0 = no precipitation and 1.0 = 100% of historical mean
- 116. MB Drought Risk: Combined Precip Temp Output
- 117. 7 Bonus… GeoParquet
- 118. Slide Title Optimize reading and analysis of published large vector dataset Goal Block Key GeoParquet reader performance demo Result Internet bandwidth and local processing limitations Structure data so you only read what you need Test case: Geoparquet is 2 - 3 X faster than other alternatives
- 119. GeoParquet ● Cloud native / cloud friendly vector data storage ● Built on & follows Parquet standards ● Column oriented ● Highly optimized for accessing very large data volumes where you need access to a few columns and geometry, such as for analysis ● Benefits from a mature set of applications, libraries & tools available for Parquet ● Supports a range of geometries ● Not spatially indexed yet
- 120. Demo
- 121. Performance: Geoparquet vs OSM, Geopackage 1 millions records, select and spatially analyze 107k water areas
- 122. GeoParquet Partitioning Nested structure with folder by feature type and separate files for each value for selected attribute
- 123. GeoParquet Partitioning Only read the features with the feature type and values you want Nested structure with folders by feature type and separate files for each value for selected attribute
- 124. Reader Local S3 Cloud -> local S3 Cloud -> FME Hosted OSM reporter* 23.2 60.4 38.1 Geopackage reporter* 1.2 102.8 14 GeoParquet reporter* 1.3 37.5 7.2 GeoParquet partitioned* 0.3 15.2 4.9 Performance: Geoparquet vs OSM, Geopackage *1 millions records, select and spatially analyze 100k water areas. Process time in seconds
- 125. ● Column oriented vector format ● Geoparquet test is 2 - 3 X faster than others ● Cloud native for vector not as easy as for raster, point cloud ● Adds requirement for appropriate cataloging ● Additional speed improvements with more attribute level partitioning ● This addresses some of the debate around geoparquet as cloud native Lessons GeoParquet
- 127. ● Start publishing now! ● Keep the processing close to the data ● Minimize traffic footprint - select just what you need ● Leverage data side filtering, microservices, lazy evaluation ● Metadata: enrich and update ● Optimization strategy: transactions volume vs data volume, response time requirements ● Test! Especially your core usage scenarios Considerations ● Heavier preprocessing, larger size required to structure and store data for optimized read ● Updates are a challenge - automation helps ● FME’s implementation based on third party libraries - collaboration for fixes, enhancements ● Newer cloud native formats: less data publicly available so far: COPC, ZARR ● Cloud optimized vector options - choice depend on use case: GeoParquet, FlatGeoBuf Integration Strategies
- 129. 9 Conclusion
- 131. Summary ● Cloud native is all about making it easy to publish data without a server, optimizing responses to web data requests: just read what you need! ● Safe’s strategy is to track and support emerging standards across a range of data types so FME users can stay ahead of evolving web technologies ● FME allows you to integrate between hybrid environments as needed ● Keep the processing close to the data ● Minimize traffic footprint - reader filtering ● Open standards enable community-wide adoption and access ● No one size fits all - know your key requirements & test!
- 132. 29+ 27K+ 128 190 20K+ years of solving data challenges FME Community members countries with FME customers organizations worldwide global partners with FME services 29+ 29K+ 128 140+ 25K+ years of solving data challenges FME Community members countries with FME customers organizations worldwide global partners with FME services 200K+ users worldwide Safe & FME
- 133. One platform, two technologies FME Form FME Flow Build and run data workflows Automate data workflows FME Flow Hosted Safe Software managed instance fme.safe.com/platform FME Enterprise Integration Platform Safe & FME
- 134. 10 Resources
- 135. Resources ● Radiant Earth Blog: Cloud Native Geospatial Solutions ● Cloud Native Geospatial Foundation ● Source Cooperative ● Chris Holmes: FOSS4G NA 2023 | Towards a Cloud Native Spatial Data Infrastructure ● Cloud Native Databases - Blog ● FME for Cloud Native Databases ● guide.cloudnativegeo.org ● Safe’s Participation in OGC Pilots
- 136. Data Sources STAC / COG: ● catalogue.dataspace.copernicus.eu/stac/ ● cmr.earthdata.nasa.gov/stac/ ● planetarycomputer.microsoft.com/catalog ● usgs.gov/landsat-missions/landsat-collection-2 ● planetarycomputer.microsoft.com/api/stac/v1/colle ctions/sentinel-2-l2a ● https://planetarycomputer.microsoft.com/api/stac/ v1/collections/nrcan-landcover ZARR: ● https://console.cloud.google.com/marketplace/pro duct/noaa-public/cmip6 COPC: ● github.com/PDAL/data/tree/master/autzen ● copc.io/#example-data
- 137. 11 Next Steps
- 138. Next Steps ● Coming: ○ Knowledge base landing page ○ Blogs ● Cloud native webinar part 2: FME deep dive focusing on newer formats and use cases: COPC, ZARR etc ● Community involvement: Cloud Native Geospatial Foundation, OGC ● Events: ○ GeoParquet Community Day in San Francisco: Jan 30 ○ Zarr Sprint in NYC: Feb 7 - 8 ● New functionality: what are your priorities?
- 139. Get our Ebook Spatial Data for the Enterprise fme.ly/gzc Guided learning experiences at your fingertips academy.safe.com FME Academy Resources Check out how-to’s & demos in the knowledge base community.safe.com /s/knowledge-base Knowledge Base Webinars Upcoming & on-demand webinars safe.com/webinars
- 140. ClaimYour Community Badge ● Get community badges for watching ● community.safe.com ● Today’s code: PCBFS Join the Community today!
- 141. 12 Q&A
- 142. ThankYou Recap of Next Steps 1 Join the FME Community 2 Contact us 3 Experience the FME Accelerator Please fill out our webinar survey