Deploying Distributed Raster Analytics · 2019-08-12 · Deploying Distributed Raster Analytics...
Transcript of Deploying Distributed Raster Analytics · 2019-08-12 · Deploying Distributed Raster Analytics...
Deploying Distributed Raster AnalyticsJie Zhang, Nirupam Dey
Aerial Photos
High-Resolution SatellitesSmall Sats
ArcGIS Has Five Core Capabilities for Imagery and Remote Sensing
Integrated and Massively Scalable
Indexes
AI / Deep Learning
Classification
Change Analysis
Raster
Processing
Functions
Analysis
Image
Analyst
Management
& Dissemination
Image Server
Temporal
Profile
Spectral Profile
Charting
Visualization &
Exploration
Full-Motion Video
Image Space
and Mensuration
Street Views
Oriented Imagery
Drone2Map
Drone Mapping
Map Production
DTM
Generation
Seamless
Orthophotos
Ortho Mapping
Stereo
Ortho Maker
Content
All Major Sensors
Drones
Analysis
Image Server
Indexes
AI / Deep Learning
Classification
Change Analysis
Raster
Processing
Functions
Analysis
Image AnalystOrtho Mapping
Image AnalystAnalytics
Apps(Excalibur)
ArcGIS Pro
AnalysisIntegrated and Massively Scalable
Exploitation & Visualization
What is Raster Analytics?
• The ArcGIS way to create and execute spatial analysis
models and image processing chains which leverage
distributed storage and analytics
• Requires: ArcGIS Enterprise + ArcGIS Image Server
ArcGIS ProWeb map
Viewer
Hosting server Raster Analysis Image Hosting
ArcGIS
Server
Image Server Image Server
Distributed Raster Data StoreArcGIS
Data StoreImagery and GIS Data
ArcGIS
ServicesGdb
Portal
Raster Analytics System Configuration
ArcGIS Data Store Setup and Hosting Server
• ArcGIS Data Store = a setup to configure relational
database on hosting server
• Hosting server is used to store hosted feature service
• Some raster analysis workflows generate hosted feature
service output
Federated Image Servers in Enterprise
• Raster Analysis
- Processing engine of the Raster Analysis tasks
- Must have a registered raster data store of file share or
cloud store.
• Image Hosting
- Dedicate to host the published image services
- Raster analytic result will be shared as image service on
this server
- Must have at least one file share type raster store
- And must have the same raster store as Raster Analysis
image server if they are no the same server
Raster Analysis and Image Hosting Server Function
• “Roles” assigned to federated image server site
• Split functions of server site to avoid resource
competition
• Server function can be assigned to any federated image
server but only one server is allowed to have the same
function
• You can have federated Image Server without “role”
Configuring ArcGIS Image server
Portal
Webmap
Cloud
storage
(Azure, AWS
…)
File share
storage
Raster
Data Store
for storing
CRFs
Federation
Federated Server 1
Hosting Server
Image Hosting Server
and
Raster Analysis Server
All-in-one
Configuring ArcGIS Image server
Portal
Webmap
Cloud
storage
(Azure, AWS
…)
File share
storage
Raster
Data Store
for storing
CRFs
Federation
Federated Server 1
Hosting Server
Federated Server 2
Image Hosting Server
and
Raster Analysis Server
Separate Image Server
from Hosting Server
Configuring ArcGIS Image server
Portal
Webmap
Cloud
storage
(Azure, AWS
…)
File share
storage
Raster
Data Store
for storing
CRFs
Federation
Federated Server 1
Hosting Server
Federated Server 2
Raster Analysis Server
Federated Server 3
Image Hosting Server
Further separation of
Raster Analysis Server
and Image Hosting Server
Raster Store
• Raster Store registered to Image Server is for storing
output imagery of Raster Analysis tasks
- Shared file system storage
- Cloud Object Storage
- Amazon S3
- Azure Blob Storage
- Alibaba Cloud
Note: register a Cloud Store first then register a Raster Store
reference to the Cloud Store
Optimized Cloud Raster Format (CRF)
• New ArcGIS Raster Dataset format for distributed read/write
- Default output imagery format for Raster Analysis tasks
- Image is split to “bundles” with optimized schema
- Support cloud storage and file system
- Published as image service
- Rendered through Raster Rendering service
Raster Rendering service
Image Service
CRF Imagery
Rendering request
readredirect
reference
Distributed Raster Analytics System Services
Geo Processing service as master
Image Service as worker
Image Service for rendering output
Raster Analysis Tools System Services
Note: Generate Raster
tool is the generic
service tool takes raster
function as input to
support various analytic
workflows
ref: http://esriurl.com/rasterfunc
Nirupam Dey
ArcGIS
Portal/Server
configuration
Nirupam Dey
Deployment in the Cloud
Deployment Platforms
• On-premise + network file shared storage
appliance
• Cloud platform + Cloud Object Storage
- Amazon Cloud Formation template
or ArcGIS Enterprise Cloud Builder CLI for AWS
- ArcGIS Cloud Builder for Azure
https://arcgisstore1071.s3.amazonaws.com/11595/docs/index.html
Nirupam Dey
ArcGIS Enterprise
Configuration and
Enterprise Builder
CLI
Federated with
Portal
S3 Storage for
Portal/Server
configuration
S3 Raster Storage
Professional Imagery /
Geospatial Analysts
Client consuming
Imagery ItemsBase Enterprise
EC2
Elastic IP for Portal
VPC
Elastic Load
Balancer
EC2
Configure Store
Client consuming
Hosted Image
Services
Image Services
Image Server
Image Services
Image Server
EC2 EC2
Image Services
Image Server
EC2
Dynamic Image
ServicesRaster Analytics
Elastic Load
Balancer
Client consuming
Dynamic Image
ServicesEnterprise + ArcGIS Server + Image server + RDS in AWS
• Used when we need to do raster analytics
• Mosaic datasets sit in Postgres ( RDS)
• All server machines are in cluster
• Config store is on a different ec2 machine
• Different Stack for Dynamic services and raster
analytic
RDS
Postgres RDS
ArcGIS Enterprise Cloud Builder for Azure
• Stand-alone app for both Desktop and Enterprise deployment
• Deployment template for Image Server
• Premium storage account recommended
https://enterprise.arcgis.com/en/server/latest/cloud/azure/install-azure-cloud-builder.htm
Special Enterprise Raster Analytics Configuration
• Ortho Maker
- Focus on drone imagery orthomapping
- Configure the elevation utility service
- Must have file share type of raster store
- Enable Ortho Maker app in your Enterprise
• Deep Learning
- Install additional deep learning python modules
- Parallelization using GPU server nodes
Deployment Best Practice
• Split Raster Analysis and Image Hosting server functions
• Adjust maximum number of processing service instances based on machine specs
• Shutdown unused services
• Configure local cache folder for each server node for
fast read/write of Cloud Raster Format (CRF)
• In cloud deployment, have data store close to your server machine
• 10:00 AM – 10:45 AM, Thursday, July 11
Image Processing Using Raster Functions
• 10:00 AM – 11:00 AM, Thursday, July 11
Using Machine Learning and Deep Learning with Imagery in ArcGIS (2nd)
• 2:30 PM - 3:30 PM, Thursday, July 11
ArcGIS Enterprise: Raster Analytics in ArcGIS Image Server (2nd)
Recommended Imagery Enterprise Demo Theatres/Workshops
Please Share Your Feedback in the App
Download the Esri
Events app and find
your event
Select the session
you attended
Scroll down to
“Survey”
Log in to access the
survey
Complete the survey
and select “Submit”