Designing Scalable Internet Platforms: Key Subsystems and Best Practices

Web Front‑End Application System

Large internet platforms require a reliable, secure, and extensible front‑end that transparently serves user requests without involving developers in server management. The system typically runs on Apache, Nginx, or Tengine, allowing multiple applications to share servers and scale by adding nodes.

Load Balancing

Load balancing can be implemented with hardware or software solutions.

Hardware load balancers (e.g., F5, A10) provide high efficiency by distributing traffic at the link level but are costly.

Software load balancers (e.g., LVS for layer‑4, HAProxy for layer‑4/7, Nginx for layer‑7) are cheaper or open‑source and sufficient for most traffic levels; many sites combine both types.

Database Cluster System

To support high reliability and massive concurrency, a MySQL‑based cluster is recommended with the following design:

Separate read and write databases, directing reads to dedicated replicas.

Use MySQL Replication to synchronize the primary (write) database to multiple slaves (read).

Deploy multiple primary servers to eliminate write bottlenecks and single points of failure.

Employ load‑balancing devices for read replicas to achieve high performance and scalability.

Separate database servers from application servers.

Apply LVS, HAProxy, or BigIP for load balancing among read replicas.

Application Caching System

Caching improves access speed, reduces database load, and enhances reliability. The most common solution for large‑scale web apps is in‑memory caching with Memcached, typically deployed on multiple nodes to avoid single‑point failures.

Improves request latency and overall server throughput.

Alleviates pressure on databases and storage clusters.

Multiple Memcached instances provide high availability and horizontal scalability.

Distributed Storage System

Large‑scale sites need storage that can handle massive volumes (e.g., photos, videos) and provide uniform access across all nodes. A high‑performance distributed storage solution is essential for both capacity and shared data access.

Distributed Server Management System

As traffic grows, traditional single‑machine management becomes insufficient. Centralized, group‑based, automated management tools like CfEngine enable batch task execution, configuration distribution, and secure communication via SSL between a CfEngine server and clients.

Code Deployment System

To automate code distribution across a load‑balanced cluster, a deployment system should support:

Virtual‑host based production servers that require no developer interaction.

Four development stages: internal development, internal testing, production testing, and production release.

Source control integration (SVN or Git) for version management.

Rsync or similar tools for efficient, scripted synchronization of code across servers.