Skip to content

HAR analysis: Launchpad initial load

Source: launchpad.har — landing on the launchpad page (control rooms list with event counts).

Page timings:

  • Content load: ~1,299 ms
  • Full load: ~4,122 ms

API calls (observed)

API Count Notes
GET /controlrooms?size=100 2× Duplicate — same request within ~80ms
GET /me/preferences 3× Different namespaces: cookiesPolicy, sidenav, whatsNew — expected
GET /controlrooms/{id}/events/summary 46× One per control room — classic N+1 pattern

N+1 pattern (critical)

  1. getControlRooms() returns 46 control rooms.
  2. For each control room, getControlRoomEventsImportanceSummary(id, dateRange) is called.
  3. Result: 1 + 46 = 47 API calls in ~1.5 seconds, all with the same date range.

All 46 event-summary calls use identical params: startDate=2026-03-09T23:02:43.233Z, endDate=2026-03-10T23:02:43.233Z. They fire sequentially (or in small parallel batches), taking ~700ms total.

Code-validated: control-rooms.service.ts — constructor calls getControlRooms() then loadEventSummariesSignal() which loops over each control room and calls getControlRoomEventsImportanceSummary().

Improvement opportunities

  1. Batch event summaries — Backend: single endpoint returning summaries for multiple control room IDs. 46 calls → 1. ~98% reduction for this flow.
  2. Fix duplicate getControlRooms — 2× → 1×.
  3. Lazy-load event counts — Show control room cards first, fetch event counts on scroll or when visible (virtualization / intersection observer).
  4. Throttle parallel requests — If batch is not feasible, limit concurrent requests (e.g. 5 at a time) to avoid connection saturation.

Estimated impact

  • Before: 51 API calls in ~2s (5 base + 46 event summaries).
  • After batch endpoint: ~6 calls — ~88% reduction.
  • After lazy-load (no batch): Same 51 calls, but spread over time; perceived load improves.