AWS Load Balancer Types: Complete Comparison Guide

Master AWS load balancing: ALB vs NLB vs CLB vs Gateway LB. Learn when to use each, architecture patterns, performance characteristics, pricing, security, and real-world use cases. For AWS architects, DevOps engineers, and cloud infrastructure teams.

🎯 Critical AWS Networking Decision: Choosing wrong load balancer type costs 2-5x more and reduces performance. This guide covers 4 AWS LB types, detailed comparison, cost analysis, performance benchmarks, and real-world deployment patterns. Used by 90%+ of AWS workloads.

📋 Complete AWS Load Balancer Guide

1. AWS Load Balancing Overview & Why It Matters
2. 4 AWS Load Balancer Types at a Glance
3. Application Load Balancer (ALB): Layer 7, HTTP/HTTPS
4. Network Load Balancer (NLB): Layer 4, Ultra-High Performance
5. Classic Load Balancer (CLB): Legacy, Simple
6. Gateway Load Balancer (GWLB): Virtual Appliances
7. Detailed Comparison: When to Use Each
8. Architecture & Setup Patterns
9. Pricing Analysis & Cost Optimization
10. Performance Benchmarks & Latency
11. Security, SSL/TLS, and Access Control
12. Real-World Use Cases & Examples

What Load Balancers Do: Distribute incoming traffic across multiple backend targets (EC2 instances, containers, Lambda). Provides single entry point (VIP), handles client requests, routes to healthy backends, detects failures, scales automatically.

Why AWS Load Balancing Matters: High availability (survive instance failures), scalability (handle traffic spikes), performance (distribute load evenly), security (SSL termination, DDoS protection), zero downtime updates (graceful shutdown of instances).

AWS Load Balancing Market: 90%+ of AWS workloads use load balancers. Industry standard. Mandatory for production. Cost: $15-75/month per load balancer (varies by type and traffic).

When Load Balancer is Required: Multiple backend instances, need high availability, traffic spikes, auto-scaling group, public-facing applications, multi-AZ deployments. Single instance? Maybe not. Multiple instances? Always.

1. Application Load Balancer (ALB): Layer 7 (Application). HTTP/HTTPS. Understands requests (URLs, hostnames, headers). Route /api/ → API, /images/ → image service. Most common (60% of deployments). Best for web applications, microservices. Cost: ~$20-30/month.

2. Network Load Balancer (NLB): Layer 4 (Transport). Ultra-high performance. Handles millions of requests per second. Low latency (<100 microseconds). For extreme performance (gaming, real-time bidding, financial trading). Cost: ~$30-40/month.

3. Classic Load Balancer (CLB): Legacy. Layer 4 + Layer 7 (hybrid). Older technology, being phased out. Still works, not recommended for new projects. Cost: ~$15-20/month.

4. Gateway Load Balancer (GWLB): New (2020). Layer 4. For virtual appliances (firewalls, intrusion detection, load testing). Niche use case. Cost: ~$20-30/month.

What ALB Does: Understands HTTP/HTTPS. Inspects request content (URL path, hostname, headers, query parameters). Routes based on rules. Example: request for api.example.com → API target group. request for images.example.com → images target group. Request for /health → health-check service.

ALB Strengths: Application-aware routing (most powerful), supports containers (ECS), microservices, host-based routing, path-based routing, hostname-based routing. Perfect for complex architectures with multiple services.

ALB Weaknesses: Layer 7 processing adds latency (~5-10ms). Can’t handle non-HTTP protocols (TCP, UDP). Performance ceiling at ~100K RPS (requests per second).

ALB Pricing: $20/month base + $0.006/LCU (Load Capacity Unit). LCU based on traffic volume, connections, processed bytes. Example: 10 million requests/month = ~$20 LCU cost.

Best For: Web applications, REST APIs, microservices, containerized workloads (ECS, EKS), content-based routing, multi-tenant applications.

What NLB Does: Extreme performance. Layer 4 (TCP/UDP level). Doesn’t inspect HTTP content—just looks at IP/port. Routes based on protocol, port, source IP. Handles non-HTTP protocols (databases, gaming, MQTT).

NLB Strengths: Ultra-low latency (~100 microseconds). Extreme throughput (1 million+ RPS). Handles any protocol. Supports static IP (important for whitelisting). Preserves source IP. Great for database load balancing, gaming servers, IoT.

NLB Weaknesses: No application-aware routing (can’t route based on URL). Simple IP/port only. More expensive than ALB (1.5x cost).

NLB Pricing: $30/month base + $0.006/NLCU (Network Load Capacity Unit). Higher base cost than ALB but similar per-capacity cost. Example: 10 million requests/month = ~$30 NLCU cost.

Best For: PeopleSoft AppServer load balancing, extreme latency requirements (gaming, financial trading), non-HTTP protocols (TCP/UDP), millions of RPS, static IP needs.

Cost Breakdown (Monthly): ALB: $20 base + ($0.006 × LCU). Example 10M requests = $20 + $10 = $30/month. NLB: $30 base + ($0.006 × NLCU). Same 10M requests = $30 + $10 = $40/month. CLB: $15 + ($0.01 per hour of data) = $15-75/month depending on traffic.

Cost Optimization Tips: (1) Use ALB for most cases (cheaper than NLB). (2) Consolidate traffic on one LB vs multiple (one large < many small). (3) Remove unused load balancers ($20/month savings per LB). (4) Use reserved capacity (12-month: 25% discount). (5) Monitor LCU usage (AWS CloudWatch).

Expensive Mistake: Using NLB for web app (10x latency requirement overkill). Cost: 2x ALB but no benefit. Common: teams choose NLB for “performance” without measuring actual need. Always measure first.

Cost-Performance Sweet Spot: ALB for most web workloads. NLB only if you’ve measured and proven <10ms latency is critical. For PeopleSoft: NLB recommended (TCP/port based routing, better performance for thick client).

Latency Measurements (Real AWS Data): ALB: 5-10ms (typical), 50ms (p99). NLB: 100-200 microseconds (typical), 5ms (p99). CLB: 1-2ms (typical), 20ms (p99).

Throughput Capacity: ALB: 100K RPS per LB (can auto-scale). NLB: 1M+ RPS per LB (extreme). CLB: 10K RPS per LB (limited). Real-world: most applications never hit limits (auto-scaling adds capacity before bottleneck).

Performance Impact: Switching ALB → NLB might save 5-10ms latency (if you’re hitting LB limits). For most applications, imperceptible. Worth it only if you measure <10ms is business requirement (gaming, financial trading, real-time bidding).

Monitoring Performance: CloudWatch metrics: TargetResponseTime (how long backend takes), RequestCount (traffic volume), ProcessedBytes (data). Use these to size LB and detect bottlenecks.

SSL/TLS Termination: Load balancer terminates HTTPS on frontend, communicates plain HTTP to backend. Benefits: CPU/RAM on backend saved (LB does encryption), easier certificate management (one cert on LB vs many on backends), speeds up communication.

Security Groups: Load balancer has security group (controls what traffic reaches LB). Backend instances have security group (should allow traffic only from LB, not from internet). Principle: defense in depth.

DDoS Protection: AWS Shield Standard (all LBs) provides basic DDoS protection. AWS Shield Advanced ($3K/month) for advanced protection. AWS WAF (Web Application Firewall) can attach to ALB for application-layer protection (SQL injection, XSS, etc.).

Access Logs: Enable on all LBs. Logs go to S3. Contains client IP, response time, server response code, bytes transmitted. Critical for troubleshooting, security audits, compliance. Minor cost (~$0.50/month for typical traffic).

Use Case 1: E-Commerce Website (ALB) Multiple services: product catalog, shopping cart, payment, admin. ALB routes /products/* to product service, /cart/* to cart service, /admin/* to admin service. Content-based routing. Scale each service independently. Typical: $40-60/month ALB cost, 100-500K requests/day.

Use Case 2: PeopleSoft on AWS (NLB) Multiple AppServers in ASG. Sticky sessions (preserve client affinity). NLB recommended for PeopleSoft (thick client). Latency sensitive. Cost: $50-80/month NLB + $500-2000/month EC2 (depending on instance size).

Use Case 3: Real-Time Gaming (NLB) Millions of concurrent players, global deployment. Ultra-low latency critical (<100ms). Each player connection maintained. NLB handles 1M+ connections. Cost: $100-200/month LB + $10K+/month EC2 for game servers.

Use Case 4: Database Load Balancing (NLB) Multiple read replicas (PostgreSQL, MySQL). NLB routes read traffic across replicas. Write goes to primary. NLB necessary because not HTTP protocol. Cost: $50-80/month NLB + database costs ($500-5000/month).

AWS Load Balancer Architecture & Optimization Services

PepperTech designs optimal AWS load balancer architectures. ALB vs NLB selection, high-availability setup, cost optimization, performance tuning, security hardening. Expert architects with 15+ years AWS infrastructure experience. Reduce costs 30-40%, improve performance, ensure reliability.