Backend and UI data gathering is now parallelized (agent #162, ui #75).

Backend (the dominant cost): the dashboard endpoint and the deployment list ran every docker and OS call serially, so latency grew with the number of deployments and services. Those independent calls now run concurrently behind bounded worker pools, so load time drops from the sum of the calls toward the slowest single one.

UI: the dashboard now fetches health, stats, and deployments in parallel, and the security overview no longer blocks the first render.

Still open here: progressive loading. The deployment list still returns in a single response once every status has resolved. The list itself is cheap (filesystem only); the per-deployment status check is the expensive part, which is what the parallelization above addresses. The next step is to split them: return the list fast without status, plus a batch statuses endpoint the UI patches in, so deployment names show immediately and status badges fill in as they resolve. Leaving this open to track that.

Sunbelt Computer Software

PL/B Language Development and Support

Uh oh!

performance: Loading deployments and the dashboard is generally too slow #154

Description

Metadata

Metadata

Assignees

Labels

Type

Fields

Projects

Milestone

Relationships

Development

Issue actions