Performance & Production Readiness

09:00 – 10:15

Set a target and capture a baseline

latency vs throughput, p95 and p99, a number, not a feeling

Takeaway A baseline captured for a real endpoint: p50, p95 and p99 recorded, and a target agreed.

10:30 – 11:45

Write a load test that resembles reality

k6 open vs closed models, ramps and soaks, test data that varies

Takeaway A load test you can rerun, whose result changes when the code changes.

12:00 – 13:00

Read a profile and name the method

Java Flight Recorder, flame graphs, blocked threads

Takeaway A profile taken under load, with the responsible method named and its share of time quantified.

14:00 – 15:15

Take a query from seconds to milliseconds

execution plans, composite indexes, what an index costs on write

Takeaway A slow query fixed with one index, the plan before and after read out loud.

15:30 – 17:00

Remove the access patterns that do not scale

N+1 in the SQL log, deep pagination, chatty calls

Takeaway An N+1 removed and a deep pagination rewritten, both with measurements.

09:00 – 10:15

See what the JVM is doing with memory

heap and GC in practice, container memory limits, heap dumps

Takeaway A memory leak reproduced and found in a heap dump, with the retaining path named.

10:30 – 11:45

Size pools and queues from evidence

connection pools, thread pools and queues, back pressure

Takeaway A pool resized from a measurement, throughput improved, and the reasoning written down.

12:00 – 13:00

Put timeouts at every boundary

timeout budgets, shedding load deliberately

Takeaway A capacity number you can state, and a service that degrades instead of collapsing.

14:00 – 15:15

Run a production readiness review

the checklist, blocking vs advisory findings

Takeaway A completed readiness review of an unfamiliar service, findings separated into blocking and advisory.

15:30 – 17:00

Prove your own improvement

before and after, writing it up

Takeaway A measured improvement on your own service, and a one-page write-up a reviewer would accept.