Gateway Load Balancer (GWLB) distributes traffic across healthy backend network virtual appliances (NVAs) to enable security defense, network analysis, and related functions. This page summarizes what GWLB supports.
Traffic support
| Category | Feature | Description |
|---|---|---|
| IP version | IPv4 | Supports IPv4 traffic access. |
| Listener | Protocol and port | Checks incoming IP packets at a specific IP address using a specified protocol and port. All protocols and ports are supported. |
| Listener | Traffic forwarding | Forwards IP traffic. |
Session persistence with hash-based scheduling
GWLB uses hash-based scheduling to ensure that packets belonging to the same flow always reach the same NVA. Three algorithms are supported:
| Algorithm | Hash inputs |
|---|---|
| Five-tuple hash | Source IP address, source port, destination IP address, destination port, IP protocol |
| Three-tuple hash | Source IP address, destination IP address, IP protocol |
| Two-tuple hash | Source IP address, destination IP address |
All three algorithms guarantee that requests with the same hash value are forwarded to the same backend server.
High availability and fault tolerance
GWLB provides several reliability mechanisms that eliminate single points of failure and keep traffic flowing when individual components go down.
Health check: GWLB performs health checks to determine the availability of backend services. When a backend server is declared unhealthy, GWLB stops forwarding requests to it and redistributes traffic to healthy servers.
Multi-zone deployment: A GWLB instance can be associated with GWLB endpoints across multiple zones. If the endpoint in one zone fails, the instance remains accessible through endpoints in other zones, ensuring business continuity and eliminating single points of failure.
Cross-zone forwarding: A GWLB instance in one zone can forward traffic to backend servers in another zone, allowing flexible deployment. A single GWLB instance can also be associated with backend servers across multiple zones. If a backend server in a given zone fails, GWLB stops sending traffic to it and redirects to servers in other zones.
Supported backend server types
GWLB supports the following resource types as backend servers:
Elastic Compute Service (ECS) instances
Elastic network interfaces (ENIs)
Elastic container instances
IP addresses
Graceful server removal with connection draining
Connection draining lets you remove a backend server without abruptly terminating active sessions. After a server is removed, GWLB keeps existing connections open for a specified connection draining timeout. When that timeout expires, if no new requests arrive during the idle connection timeout, the connections are closed and no further packets are forwarded to the server.
Monitoring
GWLB integrates with CloudMonitor to provide instance- and zone-level metrics, including connection, bandwidth, and health check metrics. Use these metrics to track the operational status of your GWLB deployment.