PostgreSQL in depth + Redis · Kafka

SELECT and INSERT are easy. At production traffic, most bottlenecks live in indexes · caches · asynchronous queues. This course frames the roles and boundaries of PostgreSQL, Redis, and Kafka.

Who it's for

• Devs who want to answer "why is this query slow?"
• People unsure whether to use Redis for caching, sessions, or locking
• Teams debating whether Kafka is warranted

What you can do afterwards

• Read EXPLAIN and tune indexes
• Run multiple PostgreSQL pools from one app
• Use Redis in its five roles (cache · session · rate limit · pub/sub · distributed lock)
• Vector search with pgvector + HNSW
• Design Kafka topics · consumers · backpressure
• Choose the right tier in a 3-layer cache (edge · Redis · PG)
• Send push notifications — token storage · failure-code handling · multi-device

Flow

From read performance to write recovery

1Measure

Locate bottlenecks with EXPLAIN and define pool boundaries.

2Search and cache

Place pgvector and layered caches according to role and invalidation policy.

3Write durability

Decide whether messaging is justified and design an idempotent pipeline.

4Recovery and delivery

Connect backup, restore, topics, and push failures through operational contracts.

The first half (1–5) targets read performance — indexes, pools, caches. The second half (6–10) targets write durability — messages, idempotency, recovery.

Steps

1. PostgreSQL deep dive — EXPLAIN · indexes — seq scan vs index · statistics
2. Orchestrating multiple pools — domain split · query helpers · transactions
3. pgvector + HNSW — vector type · operators · indexes · reproducibility
4. Five roles of Redis — cache · session · rate limit · pub/sub · distributed lock
5. 3-layer cache strategy — stale-while-revalidate · invalidation · TTL
6. Kafka — when · when not — event signatures · partitions · backpressure
7. Data pipelines — retries · idempotency — the myth of exactly-once · outbox
8. Backup · restore drills — pg_dump · PITR · monthly rehearsal
9. Kafka topic design — naming · partitions · Producer/Consumer · DLQ
10. Implementing push notifications — Firebase · token storage · failure codes · multi-device

Prerequisites — complete backend-with-spring or python-data-pipeline · basic PostgreSQL CRUD.