Mastering System Isolation: 9 Techniques to Boost High Availability

Isolation is achieved by separating systems and resources to minimize impact when problems occur, preventing an avalanche effect.

For example, if service A receives a traffic surge that consumes most resources, services B and C may become unresponsive; isolation confines the impact to A, improving overall system availability.

Isolation techniques are essential for high availability; common methods include:

1. Thread Isolation

For instance, Tomcat uses a thread pool to handle requests. If service A dominates the threads, requests to B may be delayed. Using multiple thread pools isolates workloads so a busy pool does not affect others.

2. Process Isolation

Applications often contain multiple modules (e.g., forum, transaction). When one module receives disproportionate traffic, it can monopolize resources and increase failure risk, potentially affecting the whole system. Splitting the system into separate processes provides physical isolation.

3. Cluster Isolation

When a single instance cannot meet demand, multiple instances are deployed as a cluster to improve performance. For example, a product service may run several instances, each handling a portion of the load.

Flash‑sale (seckill) services experience extreme spikes; isolating them prevents impact on other services.

4. Data‑Center Isolation

As systems grow, multiple data centers are deployed. Each data center’s services call only local services; if one center fails, DNS or load balancers can redirect traffic to another center.

5. Read/Write Isolation

Databases often start with a single instance handling both reads and writes. When scale increases, read‑write separation improves performance and ensures that a write‑side failure does not affect read operations.

Redis clusters can also use read/write isolation; reads are served from replicas, and if the primary fails, replicas continue to serve reads, with retry mechanisms across clusters.

6. Static/Dynamic Isolation

Web servers host static resources (JS, CSS) and dynamic resources (JSP, PHP). Static files can be offloaded to a CDN, reducing server load and speeding up delivery.

7. Crawler Isolation

If crawler traffic degrades system performance, it should be routed to a dedicated cluster at the load‑balancer level, protecting normal business traffic.

set $flag 0;
if ($http_user_agent ~* "spider") {
    set $flag "1";
}
if ($flag = "0") {
    # proxy to normal cluster
}
if ($flag = "1") {
    # proxy to crawler cluster
}

8. Hotspot Isolation

Flash‑sale services are hotspots and should be isolated. Read‑heavy hotspots can use multi‑level caching; write‑heavy hotspots can adopt a cache‑plus‑queue pattern.

9. Resource Isolation

Common resources include CPU, disk, and network. For example, containers with heavy disk writes may need separate disks; binding specific CPUs to Redis or Nginx improves performance. Big‑data or database clusters should be network‑isolated from application clusters.

Content compiled from "Core Technologies of Billion‑Scale Traffic Websites".