Application Load Balancer (ALB) is a fully managed Layer 7 load balancing service for HTTP, HTTPS, and QUIC workloads. ALB provides elastic scalability, advanced content-based routing, and deep integration with cloud-native services. ALB is the official cloud-native Ingress gateway for Alibaba Cloud.
If you need Layer 4 load balancing for TCP or UDP traffic, see What is Network Load Balancer (NLB)? For a comparison of all Server Load Balancer (SLB) products, see SLB instance types.
Benefits
Elastic Layer 7 processing: Distributes traffic through domains and virtual IP addresses (VIPs) with multi-tier forwarding. You can customize zone combinations and enable cross-zone auto scaling to avoid single-zone bottlenecks.
Advanced protocol support: Supports HTTP, HTTPS, QUIC, and gRPC for real-time audio and video, live streaming, gaming, and microservice applications.
Content-based routing: Routes traffic based on URL paths, HTTP headers, query strings, HTTP methods, cookies, and source IP addresses. Also supports redirects, URL rewrites, and custom HTTPS headers.
Security and reliability: Includes built-in DDoS protection and integrates with Web Application Firewall (WAF). Supports end-to-end HTTPS encryption, TLS 1.3, and pre-built and custom TLS/SSL policies.
Cloud-native integration: Integrates with Container Service for Kubernetes (ACK), Serverless App Engine (SAE), Function Compute, and open-source Kubernetes.
SSE streaming: Supports Server-Sent Events (SSE) streaming for real-time delivery of inference results from large language models.
Flexible billing: Uses Elastic IP Address (EIP) and Internet Shared Bandwidth for public network access. Adopts Load Balancer Capacity Unit (LCU) pricing optimized for elastic traffic peaks.
Performance metrics
Upgraded ALB instances
Each upgraded ALB instance is allocated three IP addresses from each specified vSwitch: one VIP for serving traffic and two local IP addresses for communicating with backend servers and performing health checks.
To ensure that ALB scales as expected, reserve at least eight IP addresses in each of the instance's vSwitches.
Performance metric | Auto-scaling upper limit |
Maximum queries per second (QPS) | 500,000 |
Maximum new connections per second (CPS) | 200,000 |
Maximum concurrent connections | 5,000,000 |
Maximum internal bandwidth | 25 Gbps |
The default Internet bandwidth of an ALB instance deployed in two zones is 400 Mbps. The actual Internet bandwidth is the sum of the bandwidth of all EIPs associated with the ALB instance.
ALB automatically scales its capacity within minutes to handle traffic changes. Consider using ALB LCU reservation in the following scenarios:
You are preparing for promotional events that generate sudden traffic spikes.
Your business experiences unpredictable traffic bursts.
You are launching or migrating a service that requires high performance from the start.
You need to maintain a guaranteed capacity for your business.
To ensure high availability, select at least two zones if the current region supports two or more. There are no extra charges for using multiple zones.
We recommend that you use your own domain name and point it to the ALB instance's domain name by using a CNAME record to enable public access to your services. This approach provides an SLA of up to 99.995%.
Non-upgraded ALB instances
Non-upgraded ALB instances have two IP modes: Dynamic IP and Static IP. The performance of an ALB instance varies depending on its IP mode.
An ALB instance's performance depends on its IP mode, not its instance edition.
Performance of a single ALB instance (example with two zones enabled):
IP mode | Maximum QPS | Maximum CPS | Maximum concurrent connections | Maximum internal bandwidth | Default Internet bandwidth |
Dynamic IP | 1,000,000 | 1,000,000 | 10,000,000 | 100 Gbps | 400 Mbps |
Static IP | 100,000 | 100,000 | 1,000,000 | 10 Gbps | 400 Mbps |
The actual Internet bandwidth of an instance is the sum of the bandwidth of all EIPs associated with it. The total peak bandwidth for all pay-by-data-transfer EIPs under a single Alibaba Cloud account in a single region cannot exceed 5 Gbps. For more information, see the bandwidth limit section in Pay-as-you-go. For more bandwidth, purchase an Internet Shared Bandwidth instance.
In a multi-zone region, an ALB instance starts with a capacity of 100,000 QPS, 100,000 CPS, and 1,000,000 concurrent connections. This initial capacity does not increase with the number of zones. Static IP instances are fixed at these limits. Dynamic IP instances automatically scale up to the maximums listed above based on traffic demands.
We recommend that you use your own domain name and point it to the ALB instance's domain name by using a CNAME record to enable public access to your services. This approach provides an SLA of up to 99.995%.
To ensure high availability, select at least two zones if the current region supports two or more. There are no extra charges for using multiple zones.
ALB components
Component | Description |
Instance | The core Layer 7 load balancer that distributes incoming traffic across backend servers. A single instance can handle up to 1,000,000 QPS. |
Listener | Checks for connection requests based on the configured protocol and port. Each instance requires at least one listener. You can configure up to 50 listeners per instance by default. |
Forwarding rule | Determines how an ALB instance routes requests to server groups based on conditions such as HTTP headers, cookies, and request methods. |
Server group | A logical group of backend servers, such as Elastic Compute Service (ECS) instances, Elastic Container Instances, or Elastic Network Interfaces (ENIs) that process requests. Server groups are independent of ALB instances and can be shared across instances. You can add up to 1,000 backend servers per group by default. |
Health check | Monitors backend server health. ALB detects unhealthy backend servers and stops routing traffic to them. Health check templates can be applied across server groups. |
ALB network types
ALB instances have two network types: Internet-facing and internal. Your selection determines whether the instance uses EIPs and Internet Shared Bandwidth instances.
Concept | Description |
VIP | The private IP address within a Virtual Private Cloud (VPC) that an ALB instance uses to distribute traffic. |
EIP | A public IP address for Internet-facing ALB instances. Internal instances do not require EIPs. For high availability, an Internet-facing instance should have at least two EIPs in different zones. |
Internet Shared Bandwidth | Internet Shared Bandwidth provides region-level bandwidth sharing. You can associate EIPs in the same region with an Internet Shared Bandwidth instance to share bandwidth and reduce costs. |
Domain name | A domain name resolvable on a public or private network to the corresponding VIP. You can use a CNAME record to map a custom domain name to the ALB instance's domain name. Starting from 00:00:00 on November 15, 2024 (UTC+8), new ALB instances use new domain names by default. The Alibaba Cloud platform does not allow you to directly access instances using the default platform domain names. Instances created before this date are not affected. |
Activate ALB
Click Create ALB to go to the ALB instance buy page.
Deploy and maintain ALB
After you register an Alibaba Cloud account, you can deploy and maintain ALB in the following ways:
ALB console: A web-based interface for creating, managing, and releasing ALB instances. For detailed instructions, see Create and manage ALB instances.
Alibaba Cloud SDK: SDKs in various programming languages, such as Java, Go, and Python.
OpenAPI Portal: Quick API retrieval, online API calls, and dynamic SDK sample code generation.
Terraform: An open source tool for managing Alibaba Cloud resources using configuration files.