Segment Statistics (I/O)

The Segment Statistics section analyzes disk and buffer cache I/O activity at the table and index level during the report period. It identifies which tables and indexes generated the most I/O and highlights those with poor cache hit ratios.

---

Summary Cards

Card	Description
Total Disk Reads	Total blocks read from disk + estimated data volume
Total Cache Hits	Total blocks served from the buffer cache + estimated data volume
Overall Cache Hit Ratio	Global cache efficiency across all tables and indexes — highlighted green if ≥ 95%
Low Cache Hit Tables	Number of tables with cache hit ratio < 80% — highlighted orange if > 0

---

I/O Activity Timeline

A line chart showing disk read and cache hit activity over the report period, allowing you to correlate I/O peaks with specific time windows.

---

Tab 1 — Top Tables by I/O

Tables ranked by total I/O volume (disk reads + cache hits), showing both heap (table data) and index I/O separately.

Column	Description
Rank	Position by total I/O
Table Name	Schema and table name
Heap Reads	Blocks read from disk for table data
Heap Hits	Blocks served from cache for table data
Heap Cache %	Cache hit ratio for table data — color-coded badge
Idx Reads	Blocks read from disk for indexes on this table
Idx Hits	Blocks served from cache for indexes
Idx Cache %	Cache hit ratio for indexes — color-coded badge
Total I/O	Total I/O volume in bytes
Efficiency	Overall I/O efficiency badge

Cache Hit Ratio Badge Severity

Cache Hit %	Badge	Meaning
≥ 95%	🟢 Green	Excellent — data mostly served from memory
90–95%	🔵 Blue	Good — acceptable cache efficiency
80–90%	🟠 Orange	Warning — notable disk I/O
< 80%	🔴 Red	Critical — heavy disk reads, investigate

---

Tab 2 — Top Indexes by I/O

Indexes ranked by total I/O volume, with per-scan I/O metrics.

Column	Description
Rank	Position by total I/O
Index Name	Schema and index name
Table	Associated table name
Disk Reads	Blocks read from disk for this index
Cache Hits	Blocks served from cache for this index
Cache %	Index cache hit ratio — color-coded badge
Scans	Number of index scans during the report period
I/O per Scan	Average blocks accessed per index scan
Total I/O	Total I/O volume in bytes
Efficiency	I/O efficiency badge

:::tip High I/O per Scan A high I/O per Scan value on an index may indicate that the index is used but inefficient — possibly a low-selectivity index, or an index on a very large table without proper statistics. Review the query using this index with EXPLAIN (ANALYZE, BUFFERS). :::

---

Tab 3 — Low Cache Hit Tables

This tab is visible only when at least one table has a cache hit ratio below 80%. It lists all affected tables with severity assessment and actionable recommendations.

Column	Description
Table Name	Schema and table name
Disk Reads	Blocks read from disk
Cache Hits	Blocks served from cache
Cache Hit %	Cache hit ratio — always shown as red badge in this tab
Total I/O	Total I/O volume
Severity	Severity level of the cache miss issue
Recommendation	Actionable recommendation for this table

---

Interpretation Guide

Low Overall Cache Hit Ratio (< 90%)

The entire working dataset does not fit in memory. Consider:

• Increasing shared_buffers in postgresql.conf
• Reviewing queries performing large sequential scans that pollute the cache
• Partitioning large tables to keep frequently accessed data hot

High Disk Reads on a Specific Table

A single table generating disproportionate disk reads is a strong indicator of:

• Missing indexes causing sequential scans
• Infrequent access patterns that cannot benefit from caching
• A table larger than shared_buffers being frequently scanned

High I/O on Indexes with Low Scan Count

An index generating high I/O with few scans may be:

• A rarely used index consuming cache space unnecessarily
• A bloated index that should be rebuilt with REINDEX CONCURRENTLY

All Tables Show Good Cache Hit Ratios

If all tables show cache hit ratios ≥ 95%, the buffer cache is well-sized for the current workload. Monitor this ratio as data volumes grow.

---

Next Steps

• Checkpoint & WAL Statistics →
• Wait Events →