Scalable Data Analytics and Visualization
with Cloud Optimized Services
Sudhir Raj Shrestha
sshrestha@esri.com
Steve Kopp
skopp@esri.com
AGU Fall Meeting, December 12, 2018 | Washington, DC
Living in Interoperable Open World
ETL
Google Earth
ENVI
Imagine
MapInfo
Ionic
GeoMedia
ArcGIS Server
ArcGIS Desktop
ArcGIS Online
ArcExplorer
Gaia
gvSig
OpenLayers
uDig
QGIS
AutoCAD
Services
Web Services
OGC
ArcGIS Pro
Platform
Open
Open
Specs
Open Science
Open
Data
Open APIOpen
Source
Open
Standards
Open
Platform
Open
Community
Proprietary
Open
ETL
Standards
Interoperable
Compliance
FOSS
Open
Architecture
mosaic dataset
geoprocessing tools
optimization
consumption
multidimensional filter
raster functions
Interactive/dynamic
scalable/extensible
dynamic web services
web maps & apps
web based analytics
Analyze ShareManage VisualizeIngest
raster types
crawl disk
link to pixels
animate over slices
sophisticated renderers
vector & scalar fields
GISworkflows that scale
Designed for the characteristics of Earth Observations scientific data
Data Service
Copyright © Esri. All rights reserved.
Cloud Optimized Formats
• Raw/Striped TIF
- Disadvantage : Sequential access
- Compression:
- None
• Tiled GeoTIF + COG
- Advantage: Tiled Access
- Compression
- None
- Lossy JPEG (8bit & 12bit)
- Lossless Deflate LZW…Generic lossless
• JPEG2000
- Advantage : Tiled Access, Higher Compression
- Disadvantage : Computationally expensive
- Compression
- Lossy
- Lossless
• MRF (Meta Raster Format)
- Advantage : Tiled Access +
Optimization
- Compression
- LERC – Controlled Lossy – Very Fast –
Optimum for higher bit depth and categorical
• CRF (Cloud Raster Format)
Advantage :
Bundles of Tiles, Multi write
- Compression
- LERC
Scalable Cloud based Implementations : High Resolution Rapid Refresh (HRRR)
NWSModel
Generation
NCO Compute
Download
S3/Blob
Upload
Public
Agencies
Create
Mosaic Dataset
Publish
Mosaic Dataset
Elastic
EC2 Instances
NCO/IDP
Internal
AWS|Azure
Auto Scaling group
ArcGIS Portal
Elastic Load
Balancer
EC2
AGS
EC2
Configure Store
Federated with
Portal
S3 Storage User
Raster Store,
zip, fgdb
Elastic IP for Portal
Professional Imagery /
Geospatial Analysts
Client consuming
Imagery Items
Client consuming
Hosted Image
Services
VPC
VPC
Imagery S3 Storage
ArcGIS ProImage Services
Image Server
Image Services
Image Server
EC2 EC2
Image Services
Image Server
EC2
Dynamic Image
Services
Raster Analytics
Elastic Load
Balancer
Client consuming
Dynamic Image
Services
Enterprise + ArcGIS Server + Image server + RDS
• Used when we need to do raster analytics
• Mosaic sit in Postgres ( RDS)
• All server machines are in cluster
• Config store is on a different ec2 machine
More info on how to do this Imageserver Deployment Image server Cloud Formation Template
More info on how to do ArcGIS enterprise Deployment ArcGIS Enterprise Cloud Formation Template
RDS
Postgres RDS
Portal : m5.2xlarge 1 machine
Image server : m5.2xlarge 3 machine
File Store : m5.xlarge 1 machine
Postgres : RDS
Load Balance for Image server stack
Elastic IP for the portal machine
All machine in same VPC
Scalable Cloud Implementation
Global Streamflow Services – A Paradigm Shift
Global
Data
PAST – Individual Hydrologic Forecasting
Global
Services
NOW – Global Hydrologic Forecasting
Current NWS River Fcst
points ~3,600
The National Water Model expands forecasting capabilities
NWM streamflow and velocity for ~
2.7 million river reaches
Current NWS AHPS points overlaid with
NWM Stream Reaches (State level scale)
Courtesy: NOAA OWP
National Water Model Web Service
Real time Flood Inundation Forecast
National Water Model + Height Above Nearest Drainage (HAND)
Compute national Height Above Nearest Drainage (HAND)
4 days of computation on 32 processors
COTS software on standard hardware
NHD Streams
10 meter NED
Elevation
Hourly download
Convert flow to depth
through rating curve
Real time Flood Forecasting
HAND Raster Map
National Water Model
Copyright © Esri. All rights reserved.
Near real time Flood Inundation Forecast
using HAND (NOAA National Water Model)
Copyright © Esri. All rights reserved.
Dynamic scientific data computing processing:
Web Map Application: HRRR
Copyright © Esri. All rights reserved.
Space-Time Ocean Data:
Web Map Application: HYCOM

Scalable Data Analytics and Visualization with Cloud Optimized Services

  • 1.
    Scalable Data Analyticsand Visualization with Cloud Optimized Services Sudhir Raj Shrestha sshrestha@esri.com Steve Kopp skopp@esri.com AGU Fall Meeting, December 12, 2018 | Washington, DC
  • 2.
    Living in InteroperableOpen World ETL Google Earth ENVI Imagine MapInfo Ionic GeoMedia ArcGIS Server ArcGIS Desktop ArcGIS Online ArcExplorer Gaia gvSig OpenLayers uDig QGIS AutoCAD Services Web Services OGC ArcGIS Pro
  • 3.
  • 4.
    mosaic dataset geoprocessing tools optimization consumption multidimensionalfilter raster functions Interactive/dynamic scalable/extensible dynamic web services web maps & apps web based analytics Analyze ShareManage VisualizeIngest raster types crawl disk link to pixels animate over slices sophisticated renderers vector & scalar fields GISworkflows that scale Designed for the characteristics of Earth Observations scientific data Data Service
  • 5.
    Copyright © Esri.All rights reserved. Cloud Optimized Formats • Raw/Striped TIF - Disadvantage : Sequential access - Compression: - None • Tiled GeoTIF + COG - Advantage: Tiled Access - Compression - None - Lossy JPEG (8bit & 12bit) - Lossless Deflate LZW…Generic lossless • JPEG2000 - Advantage : Tiled Access, Higher Compression - Disadvantage : Computationally expensive - Compression - Lossy - Lossless • MRF (Meta Raster Format) - Advantage : Tiled Access + Optimization - Compression - LERC – Controlled Lossy – Very Fast – Optimum for higher bit depth and categorical • CRF (Cloud Raster Format) Advantage : Bundles of Tiles, Multi write - Compression - LERC
  • 6.
    Scalable Cloud basedImplementations : High Resolution Rapid Refresh (HRRR) NWSModel Generation NCO Compute Download S3/Blob Upload Public Agencies Create Mosaic Dataset Publish Mosaic Dataset Elastic EC2 Instances NCO/IDP Internal AWS|Azure
  • 7.
    Auto Scaling group ArcGISPortal Elastic Load Balancer EC2 AGS EC2 Configure Store Federated with Portal S3 Storage User Raster Store, zip, fgdb Elastic IP for Portal Professional Imagery / Geospatial Analysts Client consuming Imagery Items Client consuming Hosted Image Services VPC VPC Imagery S3 Storage ArcGIS ProImage Services Image Server Image Services Image Server EC2 EC2 Image Services Image Server EC2 Dynamic Image Services Raster Analytics Elastic Load Balancer Client consuming Dynamic Image Services Enterprise + ArcGIS Server + Image server + RDS • Used when we need to do raster analytics • Mosaic sit in Postgres ( RDS) • All server machines are in cluster • Config store is on a different ec2 machine More info on how to do this Imageserver Deployment Image server Cloud Formation Template More info on how to do ArcGIS enterprise Deployment ArcGIS Enterprise Cloud Formation Template RDS Postgres RDS Portal : m5.2xlarge 1 machine Image server : m5.2xlarge 3 machine File Store : m5.xlarge 1 machine Postgres : RDS Load Balance for Image server stack Elastic IP for the portal machine All machine in same VPC Scalable Cloud Implementation
  • 8.
    Global Streamflow Services– A Paradigm Shift Global Data PAST – Individual Hydrologic Forecasting Global Services NOW – Global Hydrologic Forecasting
  • 9.
    Current NWS RiverFcst points ~3,600 The National Water Model expands forecasting capabilities NWM streamflow and velocity for ~ 2.7 million river reaches Current NWS AHPS points overlaid with NWM Stream Reaches (State level scale) Courtesy: NOAA OWP
  • 10.
  • 11.
    Real time FloodInundation Forecast National Water Model + Height Above Nearest Drainage (HAND) Compute national Height Above Nearest Drainage (HAND) 4 days of computation on 32 processors COTS software on standard hardware NHD Streams 10 meter NED Elevation Hourly download Convert flow to depth through rating curve Real time Flood Forecasting HAND Raster Map National Water Model
  • 13.
    Copyright © Esri.All rights reserved. Near real time Flood Inundation Forecast using HAND (NOAA National Water Model)
  • 14.
    Copyright © Esri.All rights reserved. Dynamic scientific data computing processing: Web Map Application: HRRR
  • 15.
    Copyright © Esri.All rights reserved. Space-Time Ocean Data: Web Map Application: HYCOM