Architecture Diagram

Jump to bottom

Matt McLoughlin edited this page May 2, 2023 · 8 revisions

Architecture Diagram for Cromwell on Azure

Step 1

Client calls Cromwell (hosted on AKS) directly using REST API, or
Client uploads the workflow and trigger file to the Azure Storage container, kicking off the Trigger Service (hosted on AKS) which calls Cromwell on Azure (hosted on AKS)

Step 2

Cromwell interprets the workflow and uses the TES backend (hosted on AKS) to execute tasks on Azure Batch
Cromwell and TES use Azure Database for PostgreSQL for their respective databases
Trigger Service and TES use Application Insights for their logs and application metrics

Step 3

Azure Batch has one node per task (allowing each task to run sandboxed in parallel)
Azure Batch nodes will pull images from ACR (or external container registries) as defined in your WDL/CDL
Azure Batch has its own Azure Storage for its lifetime

Step 4

Your completed (or failed) workflow results will be deposited in the appropriate Azure Storage container