Load Balancer Types

Bit the Chipmunk, AWS Expert published on
4 min, 603 words

When your app grows busier than a chipmunk in acorn season, you need a way to spread the traffic around — evenly, securely, and without hiccups. That’s where AWS load balancers come in. They’re not one-size-fits-all — each is tuned for specific layers, traffic patterns, and resilience needs.

In exam questions, expect to choose which load balancer best fits the requirement, not how to configure it. So let’s review what each does and what traps to avoid!


2. ⚙️ Types of Load Balancers in AWS

TypeOSI LayerBest ForExample Exam ScenarioExam Tip / Trap
Application Load Balancer (ALB)Layer 7HTTP/HTTPS, modern web apps, microservicesYou need to route based on hostnames (api.example.com vs app.example.com) or paths (/login, /checkout).ALB supports content-based routing and WebSocket — but is not aware of TCP/UDP!
Network Load Balancer (NLB)Layer 4High-performance, low-latency apps (TCP, UDP, TLS passthrough)You must handle millions of requests per second or use static IPs for whitelisted clients.Use NLB for static IPs or TLS termination with private certs — not ALB.
Gateway Load Balancer (GWLB)Layer 3Network appliances (firewalls, packet inspection)You need to insert security appliances transparently between VPCs or on ingress.GWLB = Traffic steering layer. It doesn’t terminate connections — it tunnels them (GENEVE protocol)!
Classic Load Balancer (CLB)L4/L7 (retired)Nothing -- this legacy option is no longer availableA legacy system originally used CLB and must migrate to ALB or NLB.CLB has been retired — if mentioned, the correct design choice is to migrate to ALB (Layer 7) or NLB (Layer 4). It may still appear in exam distractors.

3. 🌍 Designing for High Availability

AWS load balancers are Regional services — automatically spanning multiple Availability Zones (AZs). To design for HA:

Design RequirementSolution / Load Balancer BehaviorExam Cue
Survive AZ failureRegister targets in at least two AZs“Ensure web app remains available if one AZ goes down.”
Multi-Region HAPair with Route 53 failover routing“Need automatic DNS failover between Regions.”

💡 Exam Tip: Load balancers don’t replicate state — so for session-based apps, use sticky sessions (cookies) on ALB or session persistence via Redis or DynamoDB.


4. 📈 Designing for Scalability

Load balancers scale differently:

  • ALB/NLB scale automatically with load. No manual tuning required.
  • GWLB scales with target groups — the more appliances, the more throughput.

📘 Exam Cue:

“Traffic surges unpredictably and you need a managed service that automatically scales without pre-provisioning capacity.” ✅ Answer: ALB or NLB, not a custom EC2 proxy.

But remember: load balancer's targets do not scale automatically -- you need auto-scaling groups for that. We'll talk more about those in a later article.


5. 🛡️ Designing for Security

Each load balancer integrates differently with AWS security tools:

Security FeatureAvailable OnPurpose / Exam Cue
AWS WAFALB, CloudFront“Filter malicious HTTP requests.”
Security GroupsALB, NLB“Restrict inbound traffic.”
PrivateLinkNLB“Expose internal service privately across accounts.”
TLS TerminationALB, NLB“Centralize certificate management.”
VPC Flow LogsAll (via ENIs)“Audit network access patterns.”

💡 Exam Trap: GWLB doesn’t terminate connections or handle TLS — it simply forwards packets through security appliances.


6. 🧭 When Not to Use a Load Balancer

Sometimes, the exam wants you to recognize when not to reach for one:

ScenarioBetter Solution
Static website or cached contentCloudFront
DNS-based regional routingRoute 53 latency/geo routing
Packet filtering or IDSGWLB + Firewalls, not ALB/NLB directly

7. 🐿️ Bit’s Final Thoughts

When you see a question about “resilience,” “scaling,” or “security boundaries,” — that’s your clue it’s load-balancer territory. Just remember:

  • ALB = HTTP Smart 🧠
  • NLB = Network Fast
  • GWLB = Security Path 🛡️
  • CLB = History Lesson 📜