Evolution of System Architecture: From Single‑Machine to Distributed Solutions

Evolution of System Architecture

Architecture evolution is driven by the need to meet performance, security, continuity, and emerging technologies under specific business scenarios.

I divide the development of architecture into roughly four stages:

1. Single‑Machine Mode

During the early, rapid‑construction phase of IT, organizations built whatever systems they needed (ERP, HIS, etc.) without high‑availability requirements.

2. Dual‑Machine Hot‑Standby and Mirroring

As many systems went live, the impact of a core‑business outage became unacceptable, leading to Active‑Standby configurations where a standby machine can quickly take over.

These solutions waste standby resources, still have single‑point storage, and are costly; many products existed (RoseHA, MSCS, Symantec VCS, etc.). Storage‑level master‑slave and dual‑active solutions also appeared.

3. Multi‑Node Active‑Active (Node‑Level High Availability)

When business volume and data grew, single‑machine upgrades could no longer satisfy performance; horizontal scaling became necessary.

Technologies such as Oracle RAC, Microsoft AlwaysOn, and Moebius clusters allow multiple nodes to serve traffic simultaneously, distributing load and separating OLAP/OLTP workloads.

4. Distributed Architecture

Distributed architectures involve splitting data across multiple databases, tables, or services (horizontal/vertical partitioning, sharding, etc.) to overcome the limits of the third‑generation scaling.

The goal is to keep each shard small, simple, and low‑latency, but it requires skilled personnel and sophisticated design.

Other Technical Topics

Data replication technologies (DG, OGG, LogShipping, Replication) and infrastructure solutions (virtualization, hyper‑convergence, storage dual‑active) play supporting roles in high‑availability strategies.

How to Choose an Architecture

Decide which generation fits your needs: single‑machine for simple workloads, dual‑machine for basic HA, multi‑node active‑active for high load and separation, or full distributed systems when you have sufficient resources and expertise.

Implementation details, resource allocation, and maintenance complexity must be considered.

Summary

The specific generation is less important than selecting a solution that matches your business requirements for stability, security, efficiency, and continuity; a well‑planned architecture saves cost and avoids unnecessary risk.

Source: http://www.cnblogs.com/double-K/p/8970572.html