> ## Documentation Index
> Fetch the complete documentation index at: https://activepieces-feat-selfhost-appwebhooks.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Architecture

This page focuses on production architecture and tech specs, with three main parts:

## Components

![Architecture](https://mintlify.s3-us-west-1.amazonaws.com/activepieces-feat-selfhost-appwebhooks/resources/architecture.png)

### Redis

Redis is used as a message queue to store jobs that need processing. It's also used for locking mechanisms.

### PostgreSQL

PostgreSQL is used as the main database to store all data.

### Main Container

Activepieces is a single stateless service that can be scaled horizontally. Workers and the API server are part of the same service but can be separated by environment variables to avoid processing jobs.

### Tech Specs

Activepieces is not CPU-intensive in general, and a small instance is usually sufficient. The requirements vary depending on use cases.

| Component    | Memory (RAM) | CPU Cores | Notes                                                           |
| ------------ | ------------ | --------- | --------------------------------------------------------------- |
| PostgreSQL   | 1 GB         | 1         |                                                                 |
| Redis        | 1 GB         | 1         |                                                                 |
| Activepieces | 4 GB         | 2         | For high availability, it's better to deploy multiple machines. |

## Scaling

**Redis:** You don't need to increase the size of Redis because it only stores jobs while they're being processed. Activepieces use BullMQ, which can handle many jobs each second.

**PostgreSQL:** You also don't have to increase the size of PostgreSQL. It's only used for storing data and doesn't require significant CPU power.

**Activepieces:**

To make Activepieces handle more work, you can add more instances. For the suggested setup, set the worker concurrency in the environment variable to 20 on each machine.

With this setup, it can handle more than **20 million executions** monthly, including flow runs and background polling executions.

The main thing to consider when deciding to scale is how often it checks for new items. For example, if you have a 5-minute check, it will be executed 8,640 times in a month.