MinervaDB University: Advanced Database Training for Data & AI Leaders

Day 1 · Storage engine and optimizer internals across engines

Advanced database training begins where the data lives, on disk and in memory; this is the day the rest of the week is built on. PostgreSQL heap architecture, MVCC, visibility maps, TOAST and the planner's cost model; InnoDB redo and undo, change buffering, the adaptive hash index and MySQL's cost-based optimizer, with the 9.7 hypergraph optimizer examined on real plans; MariaDB optimizer divergence and MyRocks; SQL Server's cardinality estimator, Query Store and intelligent query processing; MongoDB WiredTiger internals and the 8.3 cost-based ranker; SAP HANA column store, delta merge and plan cache. Lab: dissect execution plans on identical schemas across all six engines and explain, precisely, why each optimizer chose differently.

Day 2 · Performance engineering and query optimization at scale

Systematic, measurement-first tuning, the core discipline of the advanced database training. Wait-event analysis, pg_stat_statements and auto_explain, performance_schema and the sys schema, Query Store forensics, HANA M_EXPENSIVE_STATEMENTS, the MongoDB profiler. Index engineering in depth: covering, partial and expression indexes, index-only scans, clustered versus heap trade-offs, columnstore for HTAP. Connection architecture with PgBouncer and ProxySQL. Lab: each participant tunes a deliberately degraded 500 GB workload and defends every change with before-and-after telemetry.

Day 3 · High availability, replication and disaster recovery engineering

Streaming and logical replication in PostgreSQL with Patroni-orchestrated failover; MySQL Group Replication and Galera semantics, certification, flow control and split-brain; SQL Server Always On Availability Groups and quorum design; MongoDB replica-set elections and write concerns; HANA System Replication modes and takeover mechanics. RPO and RTO engineering as a financial calculation. Backup theory that survives auditors: PITR, pgBackRest, XtraBackup, snapshot strategies and restore drills. Lab: forced failovers with data-loss forensics; participants measure their actual RPO, not their assumed one.

Day 4 · Cloud DBaaS architecture, economics and exit strategy

The deepest treatment of managed database services in any advanced database training programme. Aurora's decoupled storage against vanilla RDS; AlloyDB's columnar engine and read pools; Azure SQL Hyperscale; Atlas topology and search nodes; DynamoDB, Cosmos DB, Spanner and Bigtable for planet-scale patterns; ElastiCache and Memorystore; Redshift and BigQuery boundary-setting. Managed-versus-self-managed divergence maps: what each provider silently changes, restricts or lags. Cloud database FinOps: reserved-capacity arithmetic, I/O pricing traps, egress-aware DR design and negotiating leverage. Lab: build a defensible TCO model and an exit plan for a reference estate.

Day 5 · Architecture Council: capstone design review

The advanced database training capstone alumni talk about for years, and the signature MinervaDB University experience. Participants bring an anonymised architecture from their own estate, or select one of the production-derived case studies, and present it to the Architecture Council: Shiv Iyer and the Senior Principal faculty, in the same review format MinervaDB runs for Fortune 500 consulting clients. Sharding strategy, wraparound risk, HA topology, security posture, compliance boundaries and cost structure are challenged line by line. Every participant leaves with a written, prioritised architecture assessment of their design.

-- the first thing every Program 1 participant runs on their own cluster
EXPLAIN (ANALYZE, BUFFERS, SETTINGS)
SELECT o.order_id, o.total, c.segment
FROM   orders o
JOIN   customers c ON c.customer_id = o.customer_id
WHERE  o.created_at >= now() - interval '7 days'
AND    c.segment = 'enterprise';
-- Day 1 explains every line of the output; Day 2 makes it faster;
-- Day 5 asks whether the table should have been partitioned at all

The advanced database training is led by someone whose career spans more than two decades at the centre of the open-source database world: Principal Database Architect at MySQL and Sun Microsystems, senior leadership at Percona and PalominoDB, and architect of record for some of the largest PostgreSQL, MySQL and ClickHouse estates in production today. He is a prolific author on database internals, optimizer mechanics, storage engines and performance engineering, and a frequent conference speaker. Very few people have tuned as many production database systems, across as many engines, as your lead instructor.

The Senior Principal faculty are specialists in PostgreSQL internals, SQL Server Always On, MongoDB distributed systems, SAP HANA, Cassandra at petabyte scale, Redis and Valkey engineering and production vector search. They are not professional trainers reading courseware; they are the engineers enterprise clients call at two in the morning. Faculty for each cohort is selected to match the estates and priorities declared in participant intakes, and the public engineering blog offers a small taste of the depth they bring.

Request the advanced database training cohort calendar by booking an appointment or writing to the MinervaDB University desk with your preferred programme and quarter. Complete the technical intake, a short structured questionnaire covering production engines, data volumes, cloud footprint and the architecture decisions ahead of you. Seats are confirmed on intake acceptance to protect cohort seniority and peer quality.

Ten days before the cohort begins you receive lab credentials, the pre-reading dossier from the Research Library and the Zoom calendar for your cohort window. Seats are limited to fifteen per cohort and cohorts for the current quarter routinely close early; private cohorts are the preferred route for organisations sending five or more senior leaders.