Medir la tasa de aciertos de caché

Oracle

SELECT metric_name, ROUND(value, 2) AS pct
FROM v$sysmetric
WHERE metric_name = 'Buffer Cache Hit Ratio'
  AND intsize_csec = (SELECT MAX(intsize_csec) FROM v$sysmetric);

Una tasa baja no siempre es mala; interprétela en el contexto de la carga de trabajo, no de forma aislada.

PostgreSQL

SELECT ROUND(sum(blks_hit) * 100.0 / NULLIF(sum(blks_hit) + sum(blks_read), 0), 2) AS cache_hit_pct
FROM pg_stat_database;

El objetivo suele ser 99%+. Si es baja, revise shared_buffers y la caché del SO.

MySQL

SELECT ROUND((1 - (
    (SELECT variable_value FROM performance_schema.global_status WHERE variable_name='Innodb_buffer_pool_reads') /
    (SELECT variable_value FROM performance_schema.global_status WHERE variable_name='Innodb_buffer_pool_read_requests')
  )) * 100, 2) AS buffer_pool_hit_pct;

reads son lecturas desde disco, read_requests son el total de lecturas. Si es baja, aumente innodb_buffer_pool_size.

SQL Server

SELECT ROUND(
  100.0 * (SELECT cntr_value FROM sys.dm_os_performance_counters WHERE counter_name='Buffer cache hit ratio')
        / NULLIF((SELECT cntr_value FROM sys.dm_os_performance_counters WHERE counter_name='Buffer cache hit ratio base'), 0), 2) AS pct;

En SQL Server moderno, Page Life Expectancy (PLE) es un indicador más significativo de la presión de memoria.

MongoDB

var c = db.serverStatus().wiredTiger.cache;
({ configuredMB: +(c['maximum bytes configured']/1048576).toFixed(0),
   inCacheMB:    +(c['bytes currently in the cache']/1048576).toFixed(0),
   readIntoCache: c['pages read into cache'],
   evictedPages:  c['unmodified pages evicted'] })

Valores altos de 'pages read into cache' y de eviction indican que la caché de WiredTiger es demasiado pequeña.

Redis

INFO stats
# keyspace_hits / (keyspace_hits + keyspace_misses) = hit ratio

Una tasa de aciertos persistentemente baja indica que hay que revisar la configuración de TTL/eviction o el patrón de acceso.