supabase-performance-tuning（Supabase 性能 tuning）

Supabase Performance Tuning

Overview

Systematically improve Supabase query and database performance across three layers: PostgreSQL engine (indexes, query plans, materialized views), Supabase infrastructure (Supavisor connection pooling, Edge Functions, read replicas), and client SDK patterns (column selection, pagination, RPC functions). Every technique here is measurable — run EXPLAIN ANALYZE before and after to confirm the improvement.

Prerequisites

• Supabase project (local or hosted) with @supabase/supabase-js v2+ installed
• Supabase CLI installed (npx supabase --version to verify)
• Access to the SQL Editor in the Supabase Dashboard or a direct Postgres connection
• pg_stat_statements extension enabled (Step 1 covers this)

Instructions

Step 1: Diagnose — Find What Is Slow

Start every performance effort with data. Enable pg_stat_statements and run the Supabase CLI diagnostics to identify bottlenecks before optimizing.

Enable the stats extension:

CREATE EXTENSION IF NOT EXISTS pg_stat_statements;

Find the slowest queries by average execution time:

SELECT
  query,
  calls,
  mean_exec_time::numeric(10,2) AS avg_ms,
  total_exec_time::numeric(10,2) AS total_ms,
  rows
FROM pg_stat_statements
ORDER BY mean_exec_time DESC
LIMIT 10;

Check index usage and cache hit rates with the Supabase CLI:

# Which indexes are actually being used?
npx supabase inspect db index-usage

# What percentage of queries are served from cache vs disk?
npx supabase inspect db cache-hit

# Tables consuming the most space
npx supabase inspect db table-sizes

Inspect active connections for pooling issues:

SELECT state, count(*), max(age(now(), state_change)) AS max_age
FROM pg_stat_activity
WHERE datname = current_database()
GROUP BY state;

If idle connections exceed your plan's limit or active queries show high max_age, connection pooling (Step 2) and query optimization (Step 3) are the priority.

Step 2: Indexes and Query Plans

Indexes are the single highest-impact optimization. Use EXPLAIN ANALYZE to read query plans, then create targeted indexes.

Read a query plan:

EXPLAIN (ANALYZE, BUFFERS, FORMAT TEXT)
SELECT * FROM users WHERE email = 'alice@example.com';

Look for Seq Scan on large tables — that means no index is being used. After adding an index, the plan should show Index Scan or Index Only Scan.

Create a basic index:

CREATE INDEX idx_users_email ON users(email);

Create a composite index for multi-column filters:

-- Optimizes: WHERE user_id = ? AND created_at > ? ORDER BY created_at DESC
CREATE INDEX idx_orders_user_created
  ON orders(user_id, created_at DESC);

Create a partial index to cover a common filter pattern:

-- Only indexes incomplete todos — much smaller and faster than full-table index
CREATE INDEX idx_todos_user_incomplete
  ON todos(user_id, inserted_at DESC)
  WHERE is_complete = false;

Find missing indexes on foreign keys (common source of slow JOINs):

SELECT
  tc.table_name,
  kcu.column_name AS fk_column,
  'CREATE INDEX idx_' || tc.table_name || '_' || kcu.column_name
    || ' ON public.' || tc.table_name || '(' || kcu.column_name || ');' AS fix
FROM information_schema.table_constraints tc
JOIN information_schema.key_column_usage kcu
  ON tc.constraint_name = kcu.constraint_name
LEFT JOIN pg_indexes i
  ON i.tablename = tc.table_name
  AND i.indexdef LIKE '%' || kcu.column_name || '%'
WHERE tc.constraint_type = 'FOREIGN KEY'
  AND tc.table_schema = 'public'
  AND i.indexname IS NULL;

Find unused indexes (candidates for removal to reduce write overhead):

SELECT schemaname, relname, indexrelname, idx_scan
FROM pg_stat_user_indexes
WHERE idx_scan = 0 AND schemaname = 'public'
ORDER BY pg_relation_size(indexrelid) DESC;

Always use CREATE INDEX CONCURRENTLY on production tables to avoid locking writes during index creation.

Step 3: Client SDK and Infrastructure Optimization

Optimize the Supabase JS client calls, then leverage infrastructure features for scale.

**Select only needed columns — avoid select('*'):**

import { createClient } from '@supabase/supabase-js'

const supabase = createClient(
  process.env.SUPABASE_URL!,
  process.env.SUPABASE_ANON_KEY!
)

// BAD: fetches every column, large payloads
const { data } = await supabase.from('users').select('*')

// GOOD: only the columns you need
const { data } = await supabase.from('users').select('id, name, avatar_url')

Paginate with .range() instead of loading all rows:

// Page 1: rows 0-49
const { data: page1 } = await supabase
  .from('products')
  .select('id, name, price')
  .order('created_at', { ascending: false })
  .range(0, 49)

// Page 2: rows 50-99
const { data: page2 } = await supabase
  .from('products')
  .select('id, name, price')
  .order('created_at', { ascending: false })
  .range(50, 99)

Use RPC functions to push complex logic to Postgres:

-- Create a server-side function for an expensive aggregation
CREATE OR REPLACE FUNCTION get_dashboard_stats(org_id uuid)
RETURNS json AS $$
  SELECT json_build_object(
    'total_users', (SELECT count(*) FROM users WHERE organization_id = org_id),
    'active_projects', (SELECT count(*) FROM projects WHERE organization_id = org_id AND status = 'active'),
    'tasks_completed_30d', (SELECT count(*) FROM tasks t
      JOIN projects p ON p.id = t.project_id
      WHERE p.organization_id = org_id
      AND t.completed_at > now() - interval '30 days')
  );
$$ LANGUAGE sql STABLE;

// One network call instead of three separate queries
const { data } = await supabase.rpc('get_dashboard_stats', {
  org_id: 'your-org-uuid'
})

Create materialized views for expensive aggregations:

-- Precompute a leaderboard instead of recalculating on every request
CREATE MATERIALIZED VIEW leaderboard AS
SELECT
  u.id,
  u.username,
  count(t.id) AS tasks_completed,
  rank() OVER (ORDER BY count(t.id) DESC) AS rank
FROM users u
LEFT JOIN tasks t ON t.assignee_id = u.id AND t.status = 'done'
GROUP BY u.id, u.username;

-- Create an index on the materialized view
CREATE UNIQUE INDEX idx_leaderboard_user ON leaderboard(id);

-- Refresh on a schedule (e.g., via pg_cron or a cron Edge Function)
REFRESH MATERIALIZED VIEW CONCURRENTLY leaderboard;

Configure connection pooling with Supavisor:

// For serverless environments (Vercel, Netlify, Cloudflare Workers):
// Use the pooled connection string with transaction mode
// Dashboard → Settings → Database → Connection string → "Transaction mode"

// The JS SDK uses PostgREST (HTTP) which has its own pooling — no config needed.
// Direct Postgres clients (Prisma, Drizzle, pg) need the pooled string:
import { Pool } from 'pg'

const pool = new Pool({
  connectionString: 'postgres://postgres.[ref]:[pwd]@aws-0-[region].pooler.supabase.com:6543/postgres',
  max: 5,  // Keep low in serverless — Supavisor manages the upstream pool
  idleTimeoutMillis: 10000,
})

Use Edge Functions for compute-heavy operations close to data:

// supabase/functions/generate-report/index.ts
// Edge Functions run in the same region as your database — low latency
import { createClient } from 'https://esm.sh/@supabase/supabase-js@2'

Deno.serve(async (req) => {
  const supabase = createClient(
    Deno.env.get('SUPABASE_URL')!,
    Deno.env.get('SUPABASE_SERVICE_ROLE_KEY')!
  )

  // Heavy aggregation runs next to the database, not in the user's browser
  const { data } = await supabase.rpc('get_dashboard_stats', {
    org_id: (await req.json()).org_id
  })

  return new Response(JSON.stringify(data), {
    headers: { 'Content-Type': 'application/json' }
  })
})

Enable read replicas on Pro+ plans for read-heavy workloads — route analytics and reporting queries to the replica to offload the primary.

Output

After completing these steps, you will have:

• Diagnostic baseline from pg_stat_statements, index-usage, and cache-hit
• Targeted indexes on slow query columns, foreign keys, and common filter patterns
• Query plans verified with EXPLAIN ANALYZE showing Index Scan instead of Seq Scan
• Client queries optimized with column selection, pagination, and joined queries
• RPC functions and materialized views for expensive server-side aggregations
• Connection pooling configured via Supavisor for serverless deployments
• Edge Functions deployed for compute-heavy operations near the database

Error Handling

Examples

Before/after index optimization:

-- Before: 450ms, Seq Scan
EXPLAIN (ANALYZE) SELECT * FROM orders WHERE customer_id = 'abc-123';
-- Seq Scan on orders  (cost=0.00..15234.00 rows=50 width=128) (actual time=0.015..450.123 rows=50 loops=1)

CREATE INDEX idx_orders_customer ON orders(customer_id);

-- After: 0.8ms, Index Scan
EXPLAIN (ANALYZE) SELECT * FROM orders WHERE customer_id = 'abc-123';
-- Index Scan using idx_orders_customer on orders  (cost=0.42..8.44 rows=50 width=128) (actual time=0.025..0.812 rows=50 loops=1)

Client query optimization — eliminating N+1:

// BAD: N+1 — one query per project (10 projects = 11 queries)
const { data: projects } = await supabase.from('projects').select('id, name')
for (const project of projects!) {
  const { data: tasks } = await supabase
    .from('tasks').select('*').eq('project_id', project.id)
}

// GOOD: Single query with embedded join (1 query total)
const { data } = await supabase
  .from('projects')
  .select('id, name, tasks(id, title, status)')
  .eq('organization_id', orgId)

Resources

• Supabase Performance Advisor — built-in CLI diagnostics
• PostgreSQL Index Types — B-tree, GIN, GiST, and when to use each
• Connection Pooling with Supavisor — transaction vs session mode
• Supabase Edge Functions — deploy serverless functions next to your database
• Read Replicas — offload read-heavy queries on Pro+ plans
• RLS Performance Best Practices — avoid per-row subqueries

Next Steps

• For RLS policy design, see supabase-rls-policies
• For cost optimization, see supabase-cost-tuning
• For real-time subscriptions, see supabase-realtime