Selected work

Case study / 02

DASH-X

Cloud simulation platform

A cloud platform for preparing, orchestrating, and running analytical simulation workloads at scale.

Workload architectureInteractive control plane and isolated compute

The Angular client and application API form an interactive control plane. They persist scenario configuration, submit asynchronous work and report status. A separate compute plane orchestrates isolated container runs, with cloud storage holding inputs, execution state and results.

Plane 01Interactive / control plane
User workflowProjects · studies · scenarios
BrowserAngular client
Application service.NET APIConfigure · submit · monitor
Execution requestAsynchronous boundaryProgress + results
Plane 02Isolated compute plane
Control serviceWorkload orchestration
Disposable workersContainer runsStaged / parallel work
Post-runResult aggregation
Durable layerPersisted independently of workers
Scenario configurationWorkload inputsExecution stateAnalytical outputs
The browser submits and observes work through durable state; long-running compute continues independently.

01 / Workload problem

A simulation run is not a web request

Users prepare projects, studies and scenarios interactively, but the associated simulation work can be long-running and computationally intensive. It cannot depend on an open browser tab or the lifetime of a single API request.

Inputs, run state and outputs also need durable identities so a scenario can be monitored, revisited and compared after its compute process has finished.

02 / Workload architecture

A control plane submits work; a compute plane runs it

The Angular client and .NET application API form the interactive control plane. They manage scenario configuration, submission and status. A separate cloud-run service turns approved requests into staged container workloads and monitors their execution.

Scenario metadata and execution state are persisted independently of the workers. Inputs and outputs live in cloud storage, so containers can remain isolated and disposable without losing the analytical record.

  • The browser observes work; it does not host or hold it open
  • Each compute run has explicit inputs, state and outputs
  • Application and workload services can change independently

03 / Operating model

Run state is independent of the browser session

A user can submit a run, follow phase and workload status, leave the application, and return to the persisted results. The split also keeps customer-facing concerns out of the container runtime and compute-specific concerns out of the web application.