How to choose an SLB load balancer - Server Load Balancer

Product overview

Alibaba Cloud SLB is a fully managed, pay-as-you-go load balancing service for large-scale internet applications and high-concurrency scenarios such as flash sales and IoT. Unlike hardware load balancers, SLB requires no upfront investment. Unlike self-managed solutions, it provides professional operations support and free 24/7 technical assistance.

Product types

Alibaba Cloud SLB supports the following types of load balancers:

Application Load Balancer (ALB): A Layer 7 load balancer that handles up to 1 million QPS per instance. Provides HTTPS offloading and content-based routing (forwarding, redirection, and rewriting by HTTP headers, cookies, and query strings). Serves as the official cloud-native Ingress gateway on Alibaba Cloud. What is Application Load Balancer (ALB).
Network Load Balancer (NLB): A next-generation Layer 4 load balancer for IoT and high-concurrency scenarios such as large-scale device connections, messaging services, and audio/video streaming. Handles up to 100 million concurrent connections per instance. NLB features include TCPSSL offloading, new-connection rate limiting, all-port listening, and MQTTS encryption offloading. What is Network Load Balancer (NLB).
Classic Load Balancer (CLB): Supports TCP, UDP, HTTP, and HTTPS with robust Layer 4 and basic Layer 7 processing. What is Classic Load Balancer (CLB).

Feature comparison

Item	ALB	NLB	CLB
Service focus	Layer 7 processing and advanced routing features Focuses on application-layer protocols such as HTTP, HTTPS, and QUIC Optimized for application delivery	Layer 4 processing and large-scale SSL offloading Focuses on TCP, UDP, and TCPSSL protocols Optimized for network delivery	Robust Layer 4 and basic Layer 7 processing capabilities Supports TCP, UDP, HTTP, and HTTPS protocols
Architecture and performance	Built on NFV with elastic scaling Up to 1 million QPS per instance	Built on NFV, independent of physical machines, with elastic and rapid scaling Up to 100 million concurrent connections per instance	Built on a physical machine architecture Up to 1 million concurrent connections and 50,000 QPS per instance
Forwarding capabilities	Rich Layer 7 features and content-based routing HTTP header modification, redirection, rewriting, and rate limiting	Advanced Layer 4 features and TCPSSL offloading Traffic surge rate limiting, connection draining, and Anyport	Basic Layer 4 and Layer 7 capabilities Supports only domain-based or URL-based forwarding
Backend server type	Elastic Compute Service (ECS) instances elastic network interfaces elastic container instances IP addresses Function Compute	Elastic Compute Service (ECS) instances elastic network interfaces elastic container instances IP addresses	Elastic Compute Service (ECS) instances elastic network interfaces elastic container instances
O&M capabilities	Automatic scaling Scales automatically with traffic, requiring no manual intervention	Automatic scaling Scales automatically with traffic, requiring no manual intervention	Requires manual specification management for pay-by-specification instances Requires traffic peak estimation
Cloud-native integration	Cloud-native Ingress gateway Traffic splitting, traffic mirroring, canary release, and blue-green deployment	Supports integration with ACK/ASK (Version 1.24 and later)	Limited support; requires integration with Alibaba Cloud container services such as ACK or ASK
Use cases	High-performance, auto-scaling Layer 7 load balancing for internet applications High-traffic, low-latency scenarios for audio and video applications canary release and blue-green deployment for cloud-native applications	High-traffic, high-concurrency Layer 4 business scenarios Entry points for IoT services such as Internet of Vehicles (IoV) and smart home Multi-active disaster recovery and hybrid cloud ingress/egress for data centers	High-reliability Layer 4 traffic distribution for websites and applications High-concurrency, high-performance network traffic distribution Intra-city and cross-region disaster recovery

Billing comparison

The following section describes the billing for ALB, NLB, and CLB.

ALB billing

For more information about ALB resource plans, see ALB resource plans.
For pay-as-you-go ALB pricing, see ALB billing rules.

NLB billing

For pay-as-you-go NLB pricing, see NLB billing rules.

CLB billing

The following figure shows billing components for pay-as-you-go CLB instances. For pricing details, see Pay-as-you-go.

The following table compares the Load Balancer Capacity Unit (LCU) pricing and definitions for ALB, NLB, and CLB.

Service	LCU unit price Unit: USD/LCU/hour	LCU definition	References
Application Load Balancer ALB	0.007	One ALB LCU includes: 25 new connections per second 3,000 concurrent connections (sampled per minute) 1 GB of request and response data transfer per hour 1,000 rule evaluations per hour Note The number of forwarding rules, lines of AScript code, and additional certificates determine the number of rule evaluations. The free tier for each of these three metrics is 25.	LCU fee
Network Load Balancer NLB	0.005	For TCP traffic, one NLB LCU includes: 800 new TCP connections per second 100,000 concurrent TCP connections (sampled per minute) 1 GB of TCP request and response data transfer per hour For UDP traffic, one NLB LCU includes: 400 new UDP connections per second 50,000 concurrent UDP connections (sampled per minute) 1 GB of UDP request and response data transfer per hour For TCPSSL traffic, one NLB LCU includes: 50 new TCPSSL connections per second 3,000 concurrent TCPSSL connections (sampled per minute) 1 GB of TCPSSL request and response data transfer per hour	LCU fee
Classic Load Balancer CLB	0.007	For TCP traffic, one CLB LCU includes: 800 new TCP connections per second 100,000 concurrent TCP connections (sampled per minute) 1 GB of TCP request and response data transfer per hour For UDP traffic, one CLB LCU includes: 400 new UDP connections per second 50,000 concurrent UDP connections (sampled per minute) 1 GB of UDP request and response data transfer per hour For HTTP(S) traffic, one CLB LCU includes: 25 new HTTP(S) connections per second 3,000 concurrent HTTP(S) connections (sampled per minute) 1 GB of HTTP(S) request and response data transfer per hour 1,000 rule evaluations per hour Note The number of rule evaluations is determined by the number of forwarding rules. The free tier is 25.	Specification & LCU fee

Benefits

Support for multiple protocols

Supports multiple protocols for diverse application needs.

Basic: TCP and UDP (Layer 4), HTTP and HTTPS (Layer 7).
Advanced protocols:
- ALB: QUIC for audio/video and mobile apps; gRPC for microservice communication.
- NLB: TCPSSL for centralized SSL certificate management and offloading.

Multi-level disaster recovery

Multiple disaster recovery layers ensure high availability.

Health checks detect unhealthy backend servers and automatically stop forwarding traffic to them.
Multi-zone deployment: Deploy across availability zones for intra-city disaster recovery.
Session synchronization: Cluster deployment synchronizes sessions across servers, supporting hot upgrades and eliminating single points of failure.

Enhanced security and reliability

Built-in security features reduce your security overhead.

Network-layer: Protects against DDoS, SYN Flood, UDP Flood, ACK Flood, ICMP Flood, and DNS Flood attacks.
Application-layer: Integrates with Web Application Firewall (WAF) in one click for enhanced application protection.
Certificate management: Centralized management for HTTPS, QUIC, and TCPSSL certificates.

Guaranteed performance

Guaranteed traffic distribution capacity with elastic scaling.

Performance-guaranteed instances: CLB provides performance isolation between instances based on the selected specification.
Elastic capacity: ALB supports up to 1 million QPS; NLB supports up to 100 million concurrent connections and 100 Gbps bandwidth. Both auto-scale for traffic spikes.

Flexible scheduling strategies

Multiple scheduling algorithms and forwarding modes for flexible deployment.

Scheduling algorithms: Weighted round-robin, weighted least connections, source IP hash, 4-tuple hash, consistent hashing, and QUIC ID hash.
Forwarding rules: Configure rules for requests and responses to schedule traffic based on conditions and actions.

Flexible management and payment options

Multiple management channels and payment options.

Management: Console, OpenAPI, and SDKs.
Billing: Pay-as-you-go or subscription; pay by data transfer or fixed bandwidth.

Server Load Balancer:SLB product family overview