Ajuste de memoria (shared_buffers, work_mem)

Optimizing how PostgreSQL uses memory is crucial for blazing-fast performance. Think of memory as your database's short-term workspace.

Proper memory configuration means less waiting for data from slow disks and quicker query processing.

shared_buffers is one of the most important memory settings. It defines the amount of RAM PostgreSQL uses for caching data pages.

• What it does: Stores frequently accessed data blocks from tables and indexes.
• Why it matters: Reduces disk I/O, as PostgreSQL can serve data directly from RAM.

A common starting point for shared_buffers is 25% of your total system RAM. For dedicated database servers, you might go up to 40%.

• Too low? Frequent disk reads.
• Too high? Less RAM for OS/other processes, potentially causing swapping.

Remember to restart PostgreSQL after changing this setting in postgresql.conf.

work_mem is memory allocated for individual query operations like sorting (ORDER BY, GROUP BY) and hash tables (Hash Joins).

Unlike shared_buffers, this memory is private to each operation and temporary. Multiple operations in a single query, or multiple concurrent queries, can each use their own work_mem.

If a query operation exceeds its allocated work_mem, it 'spills to disk,' using temporary files. This is much slower!

You can set work_mem globally, or even per-session or per-query for specific needs. A good starting point is 4-16MB, but monitor for spills using EXPLAIN ANALYZE.

maintenance_work_mem is dedicated memory for maintenance tasks, specifically:

• VACUUM
• CREATE INDEX
• ALTER TABLE (adding columns, etc.)

These operations can be very memory-intensive. Setting this higher than work_mem is common, e.g., 64-512MB, to speed up maintenance.

effective_cache_size is not memory allocated by PostgreSQL. Instead, it tells the query planner how much memory it expects to be available for caching data.

This includes shared_buffers AND the OS file system cache. A higher value encourages the planner to use index scans over sequential scans, assuming data is likely in memory.

Most of these settings are found in your postgresql.conf file. You can also change them at runtime using ALTER SYSTEM or ALTER DATABASE/ALTER ROLE for finer control.

Remember to reload the configuration (or restart PostgreSQL for shared_buffers) after making changes to postgresql.conf.

When tuned correctly, these memory parameters work together to:

• Reduce Disk I/O: Data served from RAM is much faster.
• Speed up Queries: Sorts, joins, and aggregations complete quicker.
• Optimize Maintenance: VACUUM and index creation run faster.
• Improve Planner Decisions: Better query plans lead to better execution.

Which PostgreSQL memory parameter is primarily used by individual query operations like sorting and hashing, and can lead to "spilling to disk" if set too low?

We've explored key PostgreSQL memory parameters:

• shared_buffers: Main data cache, ~25% RAM.
• work_mem: Per-operation memory for sorts/hashes.
• maintenance_work_mem: For VACUUM, CREATE INDEX.
• effective_cache_size: A hint for the query planner.

Tuning these carefully can significantly boost your database's performance!