Elastic Network Interface Overview
Updated at：2025-10-20
Introduction to elastic network interfaces
The Baidu AI Cloud Elastic Network Interface can be migrated freely among multiple cloud servers. By attaching multiple elastic network interfaces to a single cloud server, a high-availability network solution can be achieved. Additionally, multiple private IP addresses can be associated with an elastic network interface, and each private IP address can be linked to an EIP. Once IP addresses are configured, users can design application rules to direct varied traffic flows via different source IP addresses on the network interface card, enabling deployment of a single host with multiple IP addresses.
Application scenarios
- Deploy high-availability clusters Elastic network interfaces are key components in system architecture, which meets the requirements of high-availability architecture for single instance and multiple network interface cards in the system.
- Rapid failover The traffic flow on the fault instance is quickly migrated to the standby instance by unmounting elastic network interface from the cloud host instance and then mounting it to another cloud host instance, to realize the disaster recovery deployment of services.
- Refined network management The instance can mount multiple elastic network interfaces, and elastic network interfaces can be located in different subnets and configured with the security group and ACL security access rules separately, to carry out the security access control of traffic for each elastic network interface and realize the data isolation between different businesses.
Network interface card type of cloud host
| Types | Description |
|---|---|
| Primary network interface card | Native network interface card on virtual machine instance (eth0), which is automatically generated and configured during creation of virtual machine instance. The primary network interface card can not be plugged/unplugged or migrated, and it is consistent with the lifecycle of cloud host.The BCC primary network interface card is not of elastic network interface and is not in the list of elastic network interfaces. The primary network interface card is configured on the BCC Details page. |
| Elastic network interface | Elastic network interfaces are created independently for users and can be attached to virtual machine instances. Users have the flexibility to attach elastic network interfaces during the server's creation, shutdown, or runtime. With the interface in place, users can configure system routes to direct different traffic through distinct network interface cards. |
Product limitation
- BCC instances and elastic network interfaces must reside in the same availability zone and VPC. However, they can belong to different subnets and be associated with separate security groups.
- Attaching multiple elastic network interfaces to a single instance does not increase the instance's intranet bandwidth performance.
- Each VPC allows up to 500 unmounted elastic network interfaces. Mounted elastic network interfaces do not consume quota.
- The number of IPs per network interface card must be between 1 and 40.
- The number of elastic network interfaces mounted to a cloud server depends on its instance type. For 6th generation and earlier models, the maximum number of elastic network interfaces mounted on instances is: min{CPU core count, 8}, meaning if the cloud host’s CPU core count are ≤ 8, the maximum number of elastic network interfaces mounted equals the CPU core count; if the cloud host’s CPU core count are > 8, the maximum number of elastic network interfaces mounted is 8. For 7th generation and later models, it changes depend on instance specifications. For details, refer to [Instance Specifications](BCC/Product Description/Instance specification..md).
Count of configurable IPs of network interface cards bound to the cloud server
| Memory | Count of IPs |
|---|---|
| 1G | 2 |
| (1-8]G | 8 |
| (8-32]G | 16 |
| (32-64]G | 30 |
| Greater than 64G | 40 |