Ten Database Scaling Techniques
10 entriesDatabase scaling is rarely about one technique; it is about combining the right ones at the right time. Indexing and query tuning may solve a problem for months, caching and replication can extend that further, and sharding should come only when the workload truly demands it. Denormalization helps but adds consistency and maintenance trade-offs, so base each decision on actual workload patterns, bottlenecks and access behaviour.
Tools for Scaling Databases
6 entriesSoftware commonly used to apply the techniques above.
Background and Learning
2 entriesReference reading on scaling and indexing.
Related Pages
5 entriesRelated directories on this site.
Frequently Asked Questions
What are the main database scaling techniques?
Ten common ones are indexing, vertical scaling, caching, sharding, replication, query optimization, connection pooling, vertical partitioning, denormalization and materialized views. They range from cheap, low-risk changes to heavier architectural ones.
In what order should I apply them?
Start with the cheapest and least risky: fix queries and indexes, then pool connections and add caching. Replication helps read-heavy workloads, and vertical scaling can buy time. Sharding is the most complex and should come only when the workload truly demands it.
When should I shard a database?
Consider sharding when writes or data volume exceed what one node can handle after indexing, query tuning, caching and replication have been tried. Sharding adds operational complexity, cross-shard queries and rebalancing work, so measure first.
What are the downsides of denormalization and materialized views?
Both speed up reads by storing redundant or precomputed data, which adds write complexity, storage and the risk of stale or inconsistent data. Decide based on your read and write patterns and how much staleness you can accept.
Is this page a ranking or a complete guide?
No. It is a short curated directory. Each technique links to documentation or a tool, and the right mix depends on your own workload patterns, bottlenecks and access behaviour.