CDN Multi-CDN Strategy & Failover 3 — Questions and Answers
Question 1: Which metric should be the primary signal for performance-based CDN traffic steering in a multi-CDN system?
- CDN provider's published uptime SLA percentage
- Origin server CPU utilization
- End-to-end time-to-first-byte (TTFB) measured from user locations (Correct answer)
- CDN monthly invoice cost per GB
Correct answer: End-to-end time-to-first-byte (TTFB) measured from user locations
TTFB measured from actual user locations reflects real-world CDN performance and directly correlates with user experience quality.
Question 2: A global e-commerce platform routes 70% of US traffic to CDN-A and 30% to CDN-B. CDN-A suffers a regional outage. What is the correct failover behavior if CDN-B is configured for active-active operation?
- Traffic stops until CDN-A recovers
- DNS TTL must expire before any traffic shifts to CDN-B
- The traffic steering layer redirects 100% of US traffic to CDN-B automatically (Correct answer)
- CDN-B begins serving 30% of US traffic while 70% returns 503 errors
Correct answer: The traffic steering layer redirects 100% of US traffic to CDN-B automatically
An active-active system with automated steering detects the CDN-A outage and reroutes all affected traffic to CDN-B without operator intervention.
Question 3: Which token or credential management practice is most critical when integrating multiple CDN APIs in a multi-CDN control plane?
- Store all API keys in a single shared environment variable file
- Use a secrets manager with per-CDN scoped credentials and automatic rotation (Correct answer)
- Hardcode API keys in CDN failover scripts for reliability
- Share one API key across all CDN providers to reduce management overhead
Correct answer: Use a secrets manager with per-CDN scoped credentials and automatic rotation
A secrets manager with scoped, rotatable credentials limits blast radius if one CDN API key is compromised and enforces least-privilege access.
Question 4: What is the main risk of using extremely low DNS TTLs (e.g., 5 seconds) for multi-CDN failover?
- Faster failover response times
- Increased DNS resolver query load and potential rate limiting at authoritative nameservers (Correct answer)
- Reduced cache hit rates at CDN edge nodes
- Longer time for new CDN configurations to propagate globally
Correct answer: Increased DNS resolver query load and potential rate limiting at authoritative nameservers
Very low TTLs dramatically increase DNS query frequency, which can overwhelm authoritative nameservers and DNS resolvers, creating a new point of failure.
Question 5: A SaaS company needs its multi-CDN setup to comply with GDPR by ensuring EU user data does not transit through US CDN nodes. Which capability is required?
- Content encryption at rest on all CDN edge nodes
- Geo-fencing and regional traffic isolation policies enforced at the DNS or CDN routing layer (Correct answer)
- Separate origin servers in the EU for each CDN provider
- HTTPS termination only at EU-based PoPs
Correct answer: Geo-fencing and regional traffic isolation policies enforced at the DNS or CDN routing layer
Geo-fencing policies at the routing layer ensure EU user requests are only served by EU CDN nodes, keeping personal data within required jurisdictions.
Question 6: Which CDN feature enables seamless content availability during a multi-CDN failover without requiring all traffic to hit the origin?
- Origin shield (mid-tier caching) (Correct answer)
- Edge-side includes (ESI)
- Prefetch headers
- Surrogate-Control headers
Correct answer: Origin shield (mid-tier caching)
An origin shield consolidates origin requests through a single mid-tier cache, so even during partial failovers, the shield can serve cached content to reduce origin load.
Question 7: A multi-CDN health check system reports 100% availability but users in Southeast Asia report slow load times from CDN-A. What type of monitoring gap does this scenario reveal?
- Health checks are misconfigured to check the wrong endpoints
- Binary up/down health checks cannot detect performance degradation that falls short of complete failure (Correct answer)
- CDN-A's SLA does not cover Southeast Asia PoPs
- The health check frequency is set too low
Correct answer: Binary up/down health checks cannot detect performance degradation that falls short of complete failure
Traditional health checks only detect binary availability; they cannot identify latency degradation or partial performance issues that still return 200 OK responses.
CDN Multi-CDN Strategy & Failover 3
Which metric should be the primary signal for performance-based CDN traffic steering in a multi-CDN system?