# Performance and Connection Pooling — Supabase

Source: https://www.skillbyai.com/en/supabase/p-perf

> Indexes, query plans and Supavisor.

## Tune Postgres like Postgres

Add **indexes** for columns used in filters, joins, ordering and RLS policies. Use `explain analyze` to read query plans; the dashboard's Performance Advisor and query performance reports highlight slow queries. Serverless functions open many short connections, so connect through **Supavisor**, Supabase's connection pooler: transaction mode suits serverless, session mode suits long-lived servers. Pooler hosts, ports and modes are shown in the dashboard's connect panel; check the docs for your project.

## Finding and fixing a slow query

SQL.

```sql
explain analyze
select id, title
from public.tasks
where project_id = 1 and done = false
order by id desc
limit 20;

-- a sequential scan on a large table suggests an index:
create index if not exists tasks_project_done_idx
  on public.tasks (project_id, done, id desc);

-- policy columns need indexes too
create index if not exists notes_user_id_idx on public.notes (user_id);
```

## A taxi rank instead of private cars

A pooler is a taxi rank: many passengers (requests) share a small fleet of cars (database connections) instead of each one parking their own.

**Quiz:** Why use Supavisor transaction mode for serverless functions?

- [x] Many short-lived clients share a small pool of database connections
- [ ] It disables RLS for speed
- [ ] It caches all query results forever
- [ ] It replaces indexes

*Answer:* Many short-lived clients share a small pool of database connections. Postgres connections are limited and costly.