พื้นฐานสำคัญของการขยายขนาดฐานข้อมูล

As your API grows, so does the amount of data and the number of requests to your database. A single database might struggle to keep up with the load.

Slow queries, timeouts, and even system crashes can occur, leading to a poor user experience. This is where database scaling becomes essential for maintaining performance and reliability.

There are two primary ways to scale a database:

• Vertical Scaling (Scaling Up): This means adding more resources (CPU, RAM, storage) to a single database server. It's simpler but has physical limits and creates a single point of failure.
• Horizontal Scaling (Scaling Out): This involves distributing the database load across multiple servers. It offers much greater potential for growth and is often preferred for large-scale applications.

Database replication is a technique where copies of a database are maintained on multiple servers. This setup typically involves a primary (or master) database and one or more replica (or slave) databases.

The primary database handles all write operations, and these changes are then asynchronously copied to the replicas.

The main benefit of replication is offloading read queries. Instead of all reads hitting the primary database, you can direct read requests to the replicas.

This significantly reduces the load on the primary, allowing it to focus on writes and improving overall read performance. Try running this example:

Replication offers several key advantages for scalable APIs:

• Improved Read Performance: Distributes read load across multiple servers, reducing bottlenecks.
• High Availability: If the primary database fails, a replica can be promoted to primary, minimizing downtime.
• Disaster Recovery: Replicas can be located in different geographical regions, safeguarding data.
• Reporting & Analytics: Run complex queries for reports on replicas without impacting primary database performance.

While powerful, replication has some limitations:

• Write Bottleneck: All write operations still go to the single primary database. This can become a bottleneck under very high write loads.
• Eventual Consistency: Data on replicas might be slightly out of sync with the primary for a brief period. Applications need to be designed to handle this potential delay.

For extremely high write loads or massive datasets, another strategy is often needed.

When a single primary database can no longer handle the write load, or when your dataset becomes too large for one server to store efficiently, replication alone isn't enough.

This is where sharding comes into play. Sharding is a more advanced technique to distribute both reads AND writes, and the data storage itself, across multiple independent databases.

Sharding involves breaking up a large database into smaller, more manageable pieces called shards. Each shard is an independent database instance that holds a specific subset of your total data.

Instead of one monolithic database, you have several smaller, specialized databases working in parallel. This distributes the load and storage capacity.

Choosing how to shard your data is crucial for effective scaling. Common strategies include:

• Range-based Sharding: Data is split based on a range of values (e.g., users with IDs 1-1000 on Shard A, 1001-2000 on Shard B).
• Hash-based Sharding: A hash function determines which shard a piece of data belongs to, often leading to a more even distribution.
• Directory-based Sharding: A lookup table (directory) maps data keys to their respective shards, offering flexibility but adding a lookup step.

Let's test your understanding of database scaling techniques.

We've explored essential database scaling techniques to handle growing data volumes and query loads:

• Replication creates read replicas to distribute read load, improve availability, and aid disaster recovery.
• Sharding distributes both data and write load across multiple independent databases when a single primary becomes a bottleneck.

Understanding these strategies is vital for building scalable and resilient API backends.