Why should you care?

Ops wants “orders last 5 minutes” on TV wall — rolling summary table updated every 30s beats scanning raw orders each refresh.

See it live — copy this example

Run in MySQL Workbench or the mysql CLI.

CREATE TABLE orders_per_minute (
  bucket_start DATETIME NOT NULL PRIMARY KEY,
  order_count INT NOT NULL DEFAULT 0,
  revenue_inr DECIMAL(14,2) NOT NULL DEFAULT 0
);
INSERT INTO orders_per_minute (bucket_start, order_count, revenue_inr)
VALUES (DATE_FORMAT(NOW(), '%Y-%m-%d %H:%i:00'), 1, 499.00)
ON DUPLICATE KEY UPDATE
  order_count = order_count + VALUES(order_count),
  revenue_inr = revenue_inr + VALUES(revenue_inr);
SELECT * FROM orders_per_minute
WHERE bucket_start >= NOW() - INTERVAL 15 MINUTE;

What happened?

• Upsert increments current minute bucket.
• Dashboard SELECT reads tiny rollup table — sub-10ms.
• Raw orders table untouched for live chart.

Practice next

1. Create rollup table.
2. Simulate worker INSERT ... ON DUPLICATE KEY UPDATE on each order.
3. Query last 15 minutes for chart.
4. Add avg_order_inr generated column.
5. Read rollup from replica for dashboard only.

Remember

Rollup tables for live KPIs. Upsert increments buckets atomically. Keep hot summary small; archive raw to warehouse.

DataFlow NOC screen

Grafana polls orders_per_minute every 10s during sale.

Outcome: Ops sees dip in revenue within one minute.