Elastic Load Balancing (ELB)

It distributes incoming traffic across targets (EC2, containers, IP addresses, Lambda) inside a VPC.

NGINX, HAProxy, F5 vs ELB

	NGINX / HAProxy (self-managed)	AWS ELB (managed)
Patching	self	AWS managed
Scaling	Self. Add LB instances + keepalived or DNS round-robin	AWS. Auto-scales with traffic; spans AZs by design
Health checks	Self. configure in nginx.conf / haproxy.cfg	AWS. Target group health checks in console/Terraform
Layers	NGINX(L7) HTTP; HAProxy(L4/L7)	ALB (L7), NLB (L4), or GWLB (L3 gateway for appliances)
Features	Full control: custom modules, WAF rules in nginx, Lua, etc.	ACM (TLS), WAF, Cognito auth (ALB), EKS, Auto Scaling
Cost model	For deployment on EC2	Per-hour + per LCU (capacity unit) / processed bytes

ALB, NLB, GWLB (often written “GLB”)

Type	Layer	Best for	Protocols
ELB	L4 - L7
ALB — Application Load Balancer	L7 (HTTP/HTTPS)	Web apps, microservices, path/host routing, WebSockets	HTTP, HTTPS, gRPC
NLB — Network Load Balancer	L4 (TCP/UDP)	Ultra-low latency, static IP, millions of RPS, TLS passthrough	TCP, UDP, TLS
GWLB — Gateway Load Balancer	L3 gateway	Insert third-party appliances (firewall, IDS) in the traffic path	IP packets to appliance fleet

Block diagram

flowchart LR
    Client[Client / Browser]
    DNS[Route 53 /DNS]
    GLB[Global LB]
    RLB[Regional LB]
    RLB1[Regional LB1]

    subgraph AWS Datacenter
        APIGW[API Gateway]
        ELB[ELB / ALB]
        subgraph EKS["EC2 / EKS"]
            Ingress[K8s Ingress]
            Pod1[App Svc Pod 1]
            Pod2[Pod 2]
            Pod3[Pod 3]
        end
    end

    Client -->|IP for google.com| DNS
    DNS -->|A record 1.2.3.4| Client
    Client -->|HTTP GET https:google.com| GLB
    GLB --> RLB
    GLB --> RLB1
    RLB --> APIGW
    APIGW -->|HTTP| ELB
    ELB --> Ingress
    Ingress --> Pod1
    Ingress --> Pod2
    Ingress --> Pod3