| name | zava-act2-scale |
| description | Act 2 of the Zava Lending Hyperscale workshop — Scale on Hyperscale. USE WHEN the user wants to run "Act 2", "the scale act", or to learn how Hyperscale scales: changing vCores as a minimal-downtime control-plane operation, serverless for variable load, and named read replicas for read-only scale-out. This act ships as a self-contained interactive dashboard (no Azure account needed) — guides opening scale/scale-dashboard/dashboard.html and stepping through the hypothetical 8-phase, 32→192 vCore scale-up. |
Act 2 — Scale on Hyperscale
Show how the migrated ZavaLendingDB scales for launch. This act ships as a self-contained,
interactive dashboard — no Azure account required. Narrative + concepts:
scale/README.md.
Teach the three scaling concepts first
- Changing vCores is a control-plane operation with minimal downtime. Moving the tier
(e.g. 32 → 192 vCores) is a compute reconfiguration against the same shared storage — not a
data move — so it completes in seconds with only a brief reconnect. Scale up for peak, back
down after to control cost.
- Serverless auto-scales compute within a min/max vCore range and bills for what you use —
ideal for dev/test, spiky, or unpredictable load.
- Named read replicas are independent, read-only compute nodes over the same storage —
their own endpoint and vCore tier, scaled independently, no data copying. Point reporting /
analytics / (in Act 3) vector search at a replica so reads never compete with writes.
Drive the dashboard
- Open scale/scale-dashboard/dashboard.html
— double-click or open in any modern browser. Everything is inlined; nothing to install.
- Step or play the 8 phases from the control bar (buttons 1–8, ▶ Play Progression,
Restart). The story alternates pressure (load grows on fixed vCores → CPU saturates,
latency spikes) and relief (scale vCores up online → recovers), ending with a nightly
write-only batch.
| Phase | vCores | Users | What to point out |
|---|
| 1 | 32 | 250 | Baseline |
| 2 | 32 | 500 | Pressure |
| 3 | 64 | 500 | Relief (×2) |
| 4 | 64 | 1,000 | Pressure |
| 5 | 128 | 1,000 | Relief (×2) |
| 6 | 128 | 2,000 | Pressure (peak) |
| 7 | 192 | 2,000 | Relief (top SLO) |
| 8 | 192 | — | Nightly ETL |
Two takeaways: latency returns to the same low level after every relief phase ("users never
notice the scaling"), and throughput scales near-linearly with vCores.
Full panel-by-panel guide:
scale/scale-dashboard/README.md.
Important framing
This is a hypothetical, illustrative scenario: the vCore tiers (starting at 32) and user
counts are a representative launch-day story chosen to show the scaling pattern, not a literal
continuation of Act 1 (which lands the DB on a small ~8-vCore tier). Read the numbers as an
example of how Hyperscale scales.
Next act
Continue to Act 3 (AI) via the zava-act3-ai skill.
