MySQL

MySQL 8.0's End of Life means no more security patches or bug fixes from Oracle. The migration path is to MySQL 9.7.0 LTS (released April 7, 2026) — the second LTS release after 8.4. MySQL 9.7 is the target for enterprise stability; the Innovation track (9.0-9.6) was never intended for long-term production use. Migration from 8.0 to 9.7 LTS is well-tested; the in-place upgrade path using MySQL Shell's upgrade checker utility validates compatibility before you commit. We execute these migrations with a replication-based strategy that allows rollback.

Honestly, PostgreSQL leads on most technical dimensions in 2026: 55.6% vs 40.5% developer adoption, more advanced SQL features, better extension ecosystem, stronger AI/vector capabilities via pgvector. MySQL's genuine advantages are ecosystem compatibility (WordPress, Drupal, Magento, Laravel) and broader low-cost hosting availability. For new projects without PHP ecosystem dependencies, PostgreSQL wins. For WordPress or PHP-centric applications, MySQL's compatibility advantages are real and valuable.

Oracle introduced two parallel tracks in 2023: Innovation releases (quarterly, new features, short support window — 8 months) and LTS releases (multi-year support — 5 years Premier + 3 years Extended). MySQL 8.4 was the first LTS; MySQL 9.7.0 (April 2026) is the second LTS. Production systems should run LTS releases. The Innovation track (9.0 through 9.6) was for early access to new features — not for production stability. MySQL 9.7 LTS is the right target for any production MySQL deployment in 2026.

Yes — MySQL 9.x adds a native Vector data type for storing embeddings and computing vector distances (L2, cosine, dot product). This is a meaningful addition for teams that want to add semantic search to existing MySQL applications without migrating to a different database. The honest comparison: PostgreSQL + pgvector (21K+ GitHub stars, years of production use) is more mature. MySQL's Vector type is newer and the tooling ecosystem around it is still developing. For greenfield AI applications, pgvector is the more battle-tested choice. For MySQL shops adding AI features, the native Vector type is a pragmatic upgrade path.

The standard scaling stack: (1) Connection pooling via ProxySQL — application connects to ProxySQL, which multiplexes thousands of connections over a small pool to MySQL. (2) Read replicas — add MySQL replicas, configure ProxySQL to route SELECT queries to replicas and write queries to the primary. (3) Query caching — ProxySQL's query caching layer handles repetitive read queries without hitting MySQL. (4) InnoDB buffer pool sizing — the buffer pool should fit your working set; inadequate buffer pool is the most common MySQL performance bottleneck. Beyond single-primary capacity, Vitess provides MySQL horizontal sharding used by YouTube, Slack, and GitHub at massive scale.

MariaDB is a MySQL fork created in 2009 by the MySQL founders following Oracle's acquisition. It maintains drop-in compatibility for most MySQL applications. MariaDB's advantages: fully open source (no Enterprise edition), more aggressive new features, and no Oracle governance concerns. MySQL's advantages: Oracle's investment in new features (Vector type, JavaScript stored procedures), Vitess support, and stronger tooling ecosystem from major cloud providers. For new projects: evaluate both, but don't assume MariaDB is universally superior. For existing MySQL deployments: switching to MariaDB requires testing; the surface-level compatibility doesn't guarantee behavioral equivalence for complex queries.

WordPress MySQL performance comes down to three areas: (1) Index the wp_postmeta table — add a composite index on (post_id, meta_key) or (meta_key, meta_value) depending on your query patterns; the default schema is not optimized. (2) Object caching — configure WordPress with a Redis object cache (Redis Cache Pro or W3 Total Cache). This prevents MySQL from being hit on every page load for option lookups and post metadata. (3) Query monitoring — use MySQL's slow query log with long_query_time = 1 to identify slow queries; WordPress plugins are frequent offenders. We profile WordPress MySQL performance systematically rather than applying generic optimization cargo cult.

Physical backups with Percona XtraBackup are the gold standard for large MySQL databases — it's non-blocking, hot-backup (no table locks), and restores faster than logical dumps at scale. For smaller databases (< 50GB), mysqldump or MySQL Shell's export utility are adequate. The backup strategy should include: (1) daily full backups retained 7 days, (2) point-in-time recovery via binary log archiving, (3) regular restore testing — an unverified backup is not a backup. Binary logs should be archived to S3 with sufficient retention to cover your recovery window. We configure this from day one, not after data loss.

Yes, MySQL has had a native JSON data type since 5.7. It supports JSON path expressions for querying (JSON_EXTRACT, the -> shorthand), JSON_TABLE for converting JSON to rows, and generated columns that extract JSON values for indexing. The JSON support is less capable than PostgreSQL's JSONB (which supports GIN indexes and richer operators) but handles common use cases well. MySQL 9.x extends JSON functionality alongside the new Vector type. For applications that need both relational and semi-structured data, MySQL JSON columns avoid the complexity of a separate document store.

MySQL Community Server is free (GPLv2). Infrastructure cost ranges from $0 (local dev) to ~$25-50/month (managed MySQL on DigitalOcean or RDS) to $200+/month for production-grade instances with replication. MySQL Enterprise Edition is $5,000-$50,000/year depending on features needed — rarely justified over Community + Percona Toolkit. Development cost depends on schema complexity, migration requirements, performance tuning scope, and application framework (WordPress vs Laravel vs custom). We scope specifically based on your requirements — share your project details for a precise estimate.