Load Balancing & Routing
Azure Load Balancer, Application Gateway, Front Door, Traffic Manager — choose the right load balancing service based on scope, protocol, and routing requirements.
Load balancing design
Azure has four load balancers — the right one depends on two questions:
Question 1: Global or regional? Is traffic coming from one region or worldwide?
Question 2: HTTP or TCP/UDP? Is it web traffic (HTTP/HTTPS) or general network traffic (TCP/UDP)?
Answer those two questions and you’ve narrowed to one service. This is one of the most frequently tested patterns in AZ-305.
The load balancing decision matrix
| Factor | Azure Front Door | Traffic Manager | Application Gateway | Azure Load Balancer |
|---|---|---|---|---|
| Scope | Global | Global | Regional | Regional |
| Protocol | HTTP/HTTPS | Any (DNS-based) | HTTP/HTTPS | TCP/UDP |
| OSI Layer | L7 | DNS | L7 | L4 |
| WAF | Yes (built-in) | No | Yes (optional SKU) | No |
| SSL offload | Yes | No | Yes | No |
| URL-based routing | Yes (path, header, query) | No | Yes (path, host) | No |
| Session affinity | Yes | No | Yes (cookie-based) | Yes (IP/protocol tuple) |
| Caching | Yes (edge caching) | No | No | No |
| Health probes | Yes (edge-based) | Yes (endpoint monitoring) | Yes (customisable) | Yes (TCP/HTTP probes) |
| Failover | Instant (edge-based) | DNS TTL-based (30-300 seconds) | Backend pool failover | Backend pool failover |
| Best for | Global web apps, multi-region failover, edge WAF | Global non-HTTP, multi-region DNS routing | Regional web apps, API routing, WAF | Regional VMs, internal services, non-HTTP |
The exam shortcut: Global + HTTP = Front Door. Global + non-HTTP = Traffic Manager. Regional + HTTP = Application Gateway. Regional + non-HTTP = Load Balancer. Memorise this 2x2 matrix.
Common architecture patterns
| Pattern | Services Used |
|---|---|
| Global web app | Front Door → Application Gateway → App Service/VMs |
| Global with non-HTTP | Traffic Manager → Load Balancer → VMs |
| Regional web app | Application Gateway → App Service/VMs |
| Internal services | Internal Load Balancer → VMs/VMSS |
| Multi-region HA | Front Door (primary routing) + Traffic Manager (failover) |
🏗️ Priya’s load balancing architecture:
Users → Front Door (global edge, WAF, caching)
├── East US: Application Gateway → App Service (primary)
└── West Europe: Application Gateway → App Service (secondary)- Front Door handles global routing, SSL termination, and WAF
- Application Gateway in each region provides URL-based routing to backend services
- Internal Load Balancer distributes internal API traffic between VMs
🏦 Elena’s design:
Trading clients → Traffic Manager (DNS routing, any protocol)
├── UK South: Load Balancer → Trading VMs (TCP/custom protocol)
└── West Europe: Load Balancer → Trading VMs (DR)- Trading uses a custom TCP protocol (not HTTP) → Traffic Manager + Load Balancer
- Financial data can’t use edge caching → no Front Door
Exam tip: Internal vs external Load Balancer
Azure Load Balancer has two SKUs:
- Public (External): Internet-facing — distributes traffic from the internet to VMs
- Internal: Private IP only — distributes traffic between VMs within a VNet
Common pattern: External Load Balancer for the web tier, Internal Load Balancer for the application/database tier. Internal LB keeps backend services off the internet.
Traffic Manager routing methods
| Method | How It Routes | Best For |
|---|---|---|
| Priority | Always sends to highest-priority endpoint, failover to next | Active-passive DR |
| Weighted | Distributes by weight percentage | Canary deployments, traffic splitting |
| Performance | Routes to closest (lowest latency) endpoint | Global apps, latency-sensitive |
| Geographic | Routes based on user’s geographic location | Data sovereignty, regional compliance |
| MultiValue | Returns multiple healthy endpoints | Client-side load balancing |
| Subnet | Routes based on source IP subnet | Different experiences for different networks |
Knowledge check
🏗️ GlobalTech's customer portal is deployed in two Azure regions. Users worldwide need fast page loads, the application needs WAF protection, and failover between regions should be instant (not DNS-dependent). Which load balancing service should Priya recommend?
🏦 Elena's trading platform uses a custom TCP protocol (not HTTP). It needs global routing to the closest data centre with automatic failover if a region becomes unhealthy. Which load balancing service should she recommend?
🎬 Video coming soon
Congratulations! 🎉 You’ve completed all 30 modules of the AZ-305: Designing Microsoft Azure Infrastructure Solutions study guide.
You’ve learned to design: monitoring, identity, governance, data storage, business continuity, compute, application architecture, migration, and networking — all through the lens of architecture decisions and tradeoffs.
What’s next?
- Review your flashcards and quiz scores
- Practice with scenario-based questions
- Remember: AZ-305 tests “why choose X over Y” — always explain the tradeoff