ข้อมูลแบบกระจายด้วย ETS และ Mnesia

In large-scale Erlang systems, data often needs to be shared and synchronized across multiple nodes. This is crucial for building fault-tolerant and scalable applications.

Imagine a chat application where user sessions or message queues need to be accessible even if one server fails. Distributed data stores make this possible by allowing data to live beyond a single process or node.

ETS stands for Erlang Term Storage. It's a powerful, built-in in-memory key-value store. Think of it as a super-fast hash table directly integrated into the Erlang runtime.

• Extremely fast for read/write operations.
• Can store any Erlang term (integers, atoms, lists, tuples, PIDs, etc.).
• Process-safe: multiple processes can access the same table concurrently.

ETS offers different table types, each with unique properties:

• set: The default. Each key can only have one value. Like a dictionary.
• ordered_set: Similar to set, but keys are kept sorted. Slower writes, faster range lookups.
• bag: Allows multiple objects with the same key.
• duplicate_bag: Allows multiple objects with the same key AND value.

Choosing the right type depends on your data's structure and access patterns.

Let's see how to create an ETS table and perform basic operations like inserting and looking up data. We'll use a set table for simplicity.

While ETS is incredibly fast and concurrent, it's important to remember a key limitation: ETS tables are local to the Erlang node where they are created.

• Data in an ETS table on node_A is not automatically available on node_B.
• To share ETS data across nodes, you'd need to implement custom replication or remote access mechanisms, often using message passing or RPC.

This is where Mnesia comes in to simplify distributed data management.

Mnesia is a powerful, distributed, fault-tolerant, real-time database management system built into Erlang/OTP. It leverages ETS and DETS (Disk Erlang Term Storage) for its storage.

Mnesia simplifies the challenges of managing data across multiple Erlang nodes, providing transactional guarantees and automatic replication, making it ideal for robust distributed applications.

Before using Mnesia, you need to start it and define your data schema. The schema describes your tables and their properties (e.g., what node hosts which copy).

• mnesia:start(): Initializes the Mnesia system.
• mnesia:create_schema([node()]): Creates the Mnesia directory and schema on the specified nodes. This is typically a one-time setup per cluster.
• mnesia:create_table(Name, Props): Defines a new table with its attributes and storage type.

Mnesia operations are typically performed within transactions to ensure atomicity, consistency, isolation, and durability (ACID properties). This means all operations within a transaction either succeed or fail together.

Here's how to create a table and interact with it using transactions:

Mnesia achieves distribution and fault tolerance through data replication. When creating a table, you specify where its copies should reside:

• disc_copies: Data is stored on disk and in RAM on the specified nodes. Best for persistence and fault tolerance.
• ram_copies: Data is stored only in RAM on the specified nodes. Faster access but data is lost on node crash unless a disc_copies is also present elsewhere.
• disc_only_copies: Data is stored only on disk. Slowest access but lowest RAM footprint.

Mnesia automatically handles replication and synchronization across these copies, ensuring data consistency.

In this lesson, we explored two essential tools for managing data in Erlang: ETS and Mnesia.

• ETS is a super-fast, in-memory key-value store, excellent for local caches and lookup tables. Remember, it's local to a node.
• Mnesia is Erlang's built-in distributed, fault-tolerant database. It offers transactional safety, schema definition, and automatic data replication across nodes, built upon ETS.

Understanding when to use each is key to building resilient and scalable Erlang applications.