Designing High‑Availability Service Paths: Smart DNS, Routing, CDN & GSLB Explained

Background

Service availability is a key metric for platform quality. Even when high‑availability clusters protect against server failures, the path from a user to the service can still suffer from latency, outages, or sub‑optimal routing.

Basic DNS Access

Web services use TCP/IP, which requires numeric IP addresses. DNS translates human‑readable domain names to IPs. The typical resolution flow is shown in the diagram.

This simple approach has two major drawbacks:

Single‑point failure: if the server or the IDC link fails, users experience downtime until the IP or route change propagates, delayed by DNS TTL.

Lack of optimal routing: users cannot be directed based on ISP, geographic region, or network congestion.

Optimization Techniques

Four common solutions address these drawbacks:

Smart DNS

Smart Routing

CDN multi‑point access

Global Server Load Balancing (GSLB)

Smart DNS

Many DNS providers support multiple A records for a host (DNS round‑robin). The resolver returns one of the IPs, enabling simple load‑balancing and ISP‑aware policies (e.g., China Telecom users receive a Telecom‑origin IP). No extra cost is required.

Limitations:

TTL delay when IPs change.

Very simple load‑balancing; cannot consider geographic distribution or fine‑grained policies.

Smart Routing

Smart Routing moves the decision point to the application layer. Clients query a routing service or SDK, which selects an endpoint based on response time, server load, region, etc. This is suitable for mobile apps or internal service calls.

Drawbacks:

Requires code changes in the client.

Not applicable to generic PC‑side domain access.

Complex policy configuration can cause misrouting.

CDN Multi‑point Access

In this model the domain’s C‑record points to a CDN edge node. The CDN’s routing engine selects the optimal entry point for the user and forwards the request to the origin server if needed. If the content is cached, the CDN serves it directly.

Pros and cons:

Pros: no client‑side code changes; leverages mature CDN infrastructure.

Cons: vendor opacity makes full‑stack troubleshooting difficult; additional CDN traffic incurs cost.

Global Server Load Balancing (GSLB)

GSLB distributes traffic across geographically dispersed servers, directing users to the nearest or most performant endpoint. DNS C‑records point to a Global Traffic Manager (GTM) that returns the optimal IP based on predefined policies such as latency, health checks, or geographic rules.

Typical deployment uses a commercial appliance (e.g., F5 GTM) and requires careful policy configuration.

Practical Guidance

To identify which technique a service uses, run nslookup <domain> and examine the returned records. Multiple A records suggest Smart DNS; a C‑record pointing to a CDN or GTM indicates CDN multi‑point access or GSLB respectively.

Conclusion

Ensuring high availability and optimal user paths involves trade‑offs:

Smart DNS – low cost, partial mitigation, limited granularity.

Smart Routing – fine‑grained control but requires client changes.

CDN multi‑point access – leverages CDN routing, introduces vendor lock‑in and extra traffic cost.

GSLB – provides global load distribution with higher operational complexity and cost.