Más contenido relacionado La actualidad más candente (20) Similar a Analyze Slide Images and Process Phenotypic Assays at Scale on AWS (CMP358) - AWS re:Invent 2018 (20) Más de Amazon Web Services (20) Analyze Slide Images and Process Phenotypic Assays at Scale on AWS (CMP358) - AWS re:Invent 20182. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Nick Whalen
Principal Engineer
Novartis Institutes for BioMedical Research
Gene Ting
Solution Architect
Amazon Web Services
Analyzing Slide Images and Processing
Phenotypic Assays at Scale on AWS
C M P 3 5 8
3. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Agenda
AWS services walkthrough
HCSIA overview
Compute environment design
Post-processing workflow
Architecting for resilience
Final thoughts
4. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AWS Batch
Managed
No software to install or
servers to manage.
AWS Batch provisions,
manages, and scales your
infrastructure
Integrated with AWS
Natively integrated with the
AWS products and services,
AWS Batch jobs can easily
and securely interact with
services such as Amazon
Simple Storage Service
(Amazon S3), Amazon
DynamoDB, and Amazon
Rekognition
Cost-optimized resource
provisioning
AWS Batch automatically
provisions compute resources
tailored to the needs of your
jobs using Amazon Elastic
Compute Cloud (Amazon
EC2) and Spot Instances
5. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AWS Batch key concepts
• Jobs
• Job definition
• Job queue
• Scheduler
• Compute environments
6. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AWS Step Functions
Manages the logic
of your application.
This removes extra
code that may be
repeated in your
microservices and
functions
Write less code
AWS Step Functions
manage state,
checkpoints, and
restarts to make
sure your application
executes in order
and as expected
Improve
resiliency and scale
Easy to connect and
coordinate
distributed
components and
microservices to
quickly create apps
Build and update
apps quickly
7. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon EC2 Spot Instances
Low cost Faster results Easy access Resource
flexibility
Spare EC2 capacity that AWS can reclaim with two minutes of notice
8. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
About NIBR
• The Novartis Institutes for BioMedical Research (NIBR) is the innovation
engine of Novartis
• We collaborate across scientific and organizational boundaries, with a
focus on new technologies that have the potential to help produce
therapeutic breakthroughs for patients
• Six research campuses across the globe
• 200+ projects in clinical pipeline
• 500+ clinical trials in progress
9. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
High-content screening image analysis
• Scientists need a user-friendly means to manage batch image analyses
• HCSIA empowers scientists to directly run image analyses without
depending on cluster experts or custom scripts
• HCSIA provides for faster assay development and execution with
more focus on the science rather than tools
10. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
plates
wells
11. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
12. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Architecture overview
Update job
status
Check job
status
Amazon
Aurora job
tracker*
Check job
scheduler
HCS images
Job
completion
queue
Image
analytics
results
Notify
scientists
UI and web
services
Profiler workers
HPC cluster
Users
On-premise
Post process and
merge workflow
HCSIA VPC
13. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Acquire compute resources at scale
CE1
CE2
CE3
CE4
Max vCPUs: 30000
Min vCPUs: 0
Desired vCPUs: 0
Optimal
Max vCPUs: 25984
Min vCPUs: 0
Desired vCPUs: 0
m4.16xlarge
Max vCPUs: 25984
Min vCPUs: 0
Desired vCPUs: 0
r4.8xlarge
Max vCPUs: 26000
Min vCPUs: 0
Desired vCPUs: 0
m4.10xlarge
Spot Fleet
CE1
CE2
CE3
CE4
14. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Analytics results post process and merge
15. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Merge only/merge and post process
16. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Architecture overview – job tracking
Update job
status
Check job
status
Aurora job
tracker*
Check job
scheduler
HCS images
Job
completion
queue
Image
analytics
results
Notify
scientists
UI endpoint
Profiler workers
HPC cluster
Users
On-premise
Post process and
merge workflow
HCSIA VPC
Job completion tracking
17. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Architecting against job failures
• Application failures
• Bad data
• Incorrect resource requirements
• Application bugs
• Infrastructure failures
• Disk failures
• Instance failures
• Spot reclamation
18. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Failure handling
Failure
analyzer
Job resubmit
worker
Exception
queue
Task failure
event
Fatal
exception
notification
19. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Final thoughts – Step Functions
• Be confident with state machines - use a minimal amount of steps
• Use Amazon S3 to persist and iterate over large data sets and pass
object keys
• Extract certain business functionality by state machine
20. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Final thoughts – Error handling
• Understand possible exceptions in each step
• On any given step, determine if the state machine should stop
execution or continue
• Use exponential back-off and retry
• Catch exceptions
21. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Final thoughts – Know the limits
• Ensure sufficient EBS IOPs for Docker hosts
• Iterate through large data sets using for loop or
iterator
• Avoid exceeding the maximum number of history
events in an execution (25000):
22. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
23. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Related breakouts
Wednesday, Nov. 28
High Performance Computing on AWS: Driving Innovation without Infrastructure Constraints
3:15 p.m.–4:15 p.m. | Aria East, Plaza Level, Orovada 2
Wednesday, Nov. 28
Optimizing Risk Analysis with Grid Computing on AWS
1:00 p.m.–2:00 p.m. | Venetian, Level 4, Lando 4305
Wednesday, Nov. 28
Setting Up Your First HPC Cluster on AWS
11:30 a.m.–12:30 p.m. | Mirage, Grand Ballroom B, Table 5
24. Thank you!
© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Gene Ting
geneting@amazon.com
25. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.