What's changed: Deepened AZ-305 Chapter 3 (added comparison tables, scenarios, FAQs, exam traps, deep-dive paragraphs to each section; localized figures to Japanese)
3.2Designing High Availability
Understand availability design—availability zones/availability sets, load balancing, zone-redundant services, and SLAs. Eliminate single points of failure and meet required availability.
High availability comes from redundancy. Spread across availability zones and use load balancing and zone-redundant services to remove single points of failure.
3.2.1Means of high availability
- Availability zones: spread across physically separate DCs in a region to survive zone failures.
- Availability sets: spread across fault/update domains within a zone (rack/maintenance failures).
- Load balancing: Load Balancer (L4) / Application Gateway (L7) / Front Door·Traffic Manager (global).
- Zone-redundant services/SLA: make managed services zone-redundant to meet SLA requirements.
Common on AZ-305: survive zone failure = availability zones, rack/maintenance within a zone = availability sets, L4 = Load Balancer, L7 = Application Gateway, global = Front Door/Traffic Manager. Without placing multiple VMs in zones or sets, you only get the single-VM SLA.
Application Gateway is L7 (WAF-capable), Front Door is global L7 + CDN/WAF, and Traffic Manager is DNS-based global routing.
AZ-305 availability design probes "which redundancy and load balancing meet the target SLA." Availability zones are physically independent DC groups within a region (separate power/cooling/network); spreading VMs or zone-redundant services across zones survives a zone (DC) failure. Availability sets spread within a single datacenter across fault domains (FD = power/rack units) and update domains (UD = sequential maintenance units), guarding against rack failures and planned maintenance—they cover a different failure domain, and are insufficient for a zone failure. Choose load balancing by OSI layer and scope: Load Balancer (L4, in-region TCP/UDP), Application Gateway (L7, path-based/WAF/SSL offload), Front Door (global L7 + CDN + WAF + fast failover), and Traffic Manager (DNS-based global routing: priority/weighted/geographic/performance). SLAs reach high values only with redundancy (a single VM is limited; zone spread or multiple instances improve it), using each service’s zone-redundant options (zone-redundant Application Gateway v2, zone-spread VMSS, zone-redundant SQL). The key is to clarify the failure domain to protect (node/rack/zone/region) and pair the matching redundancy with the right-layer load balancing.
| Load balancer | Layer/scope | Use |
|---|---|---|
| Load Balancer | L4, in-region | Distribute TCP/UDP |
| Application Gateway | L7, in-region | Path-based/WAF/SSL offload |
| Front Door | L7, global | CDN/WAF + fast failover |
| Traffic Manager | DNS, global | Priority/weight/geo/perf routing |
Scenario: Make a web app keep running even if one DC (zone) fails, with HTTP path-based routing and WAF. → Spread VMs/VMSS across multiple availability zones to survive a zone failure, and front it with a zone-redundant Application Gateway v2 (WAF enabled) for L7 routing. To extend across regions, add Front Door (global L7 + fast failover) at the very front.
FAQ: Availability zones or availability sets? To survive a zone (whole-DC) failure, use availability zones (spread across zones)—the default when the service supports zones. Availability sets are the older construct guarding against rack failures and planned maintenance within a single datacenter. For higher SLA and zone-failure resilience, choose availability zones.
Exam trap: Trying to satisfy a "survive a zone failure" requirement with availability sets alone is wrong. Availability sets spread fault/update domains within a single DC; a zone (whole-DC) failure needs availability zones. Also, a single VM cannot achieve a high SLA—it improves only when multiple instances are placed across zones/sets.
3.2.2Section summary
- Redundancy = availability zones (zone failure) / availability sets (rack)
- Distribution = LB (L4)/App Gateway (L7)/Front Door·Traffic Manager (global)
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. You want service to continue even if one datacenter (zone) fails. How do you place VMs?
Q2. You want L7 (HTTP/HTTPS) load balancing with path-based routing and WAF. What?
Q3. You want global optimal routing and fast delivery to endpoints in multiple regions. What?