ERIs-Elastic Compute Service(ECS)-阿里云帮助中心

An ERI is an ENI with the eRDMA feature enabled, used to provide RDMA network communication for ECS instances.

Overview

eRDMA is a low-latency, high-throughput RDMA network service from Alibaba Cloud. To use eRDMA on an ECS instance, bind ERIs to the instance. See Overview.

ERIs are supported only on specific instance families. See Instance family.

To query the maximum number of ERIs for an instance type, call the DescribeInstanceTypes operation and check the EriQuantity parameter. A value of 0 means the instance type does not support ERIs.

When binding multiple ERIs to an instance, consider the maximum QP count. See Maximum number of QPs.

Create an ERI

Create an ERI when you create an ECS instance

If the instance type supports eRDMA, enable the ERI feature for ENIs when you create an ECS instance. The configuration varies by instance type.

For enterprise-level instances, see During instance creation in "Configure eRDMA on an enterprise-level instance".
For GPU-accelerated instances, see Enable on GPU-accelerated instance.

Separately create an ERI

When you create a secondary ENI, turn on eRDMA Interface. The ERI shares the IP addresses and security group rules of the ENI. See Create and use an ENI.

Call an API operation to create an ERI

Call the CreateNetworkInterface operation and set NetworkInterfaceTrafficMode to HighPerformance to enable the ERI feature.

To bind, unbind, or modify ERIs, see Manage ENIs.

View ERIs

View ERIs in the ECS console

Check whether the ERI feature is enabled for an ENI in the ECS console.

Log in to the ECS console.
In the left-side navigation pane, choose Instances & Images > Instance.
Find the target ECS instance and click the instance ID.
Click the ENIs tab.
If flexible RDMA interface is displayed in the ENI Type column, the ERI feature is enabled for the ENI.

Call an API operation to view ERIs

Call the DescribeNetworkInterfaceAttribute operation and check the NetworkInterfaceTrafficMode parameter to determine the ERI status.

HighPerformance: ERI is enabled. eRDMA communication mode is used.
Standard: ERI is disabled. TCP communication mode is used.

View ERIs from within an ECS instance

After enabling the ERI feature for ENIs, connect to the instance to verify the ERIs.

Connect to the ECS instance.
See Log on to a Linux instance using Workbench.
View RDMA devices:
```
ibv_devices
```
The returned device name indicates an ERI.
If multiple ERI-enabled ENIs are bound to the instance, the output lists all devices.

Run ibv_devinfo to view ERI details such as port status and queue count.
If no output is returned, the eRDMA driver may be installed incorrectly or the ERI may be misconfigured. See Verify eRDMA configuration.

Enable or disable the ERI feature for an existing ENI

Call the ModifyNetworkInterfaceAttribute operation and set NetworkInterfaceTrafficMode (under NetworkInterfaceTrafficConfig) to HighPerformance to enable the ERI feature.

HighPerformance: ERI is enabled. eRDMA communication mode is used.
Standard: ERI is disabled. TCP communication mode is used.

Note

You cannot enable the ERI feature for an existing ENI in the ECS console.

Maximum number of QPs

A QP is a basic communication entity in RDMA. It consists of a Send Queue (SQ) and a Receive Queue (RQ). The QP is used to manage the data sent and received.

The maximum QP count of an RDMA device or ERI determines the maximum concurrent RDMA connections and affects network scalability.

Maximum number of QPs supported by an ECS instance

The maximum QP count varies by instance type. When multiple ERIs are bound, the instance's QP limit equals the sum of QPs across all ERIs. Call the DescribeInstanceTypes operation to query the maximum number of QPs for a specific instance type.

Enterprise-level CPU-based instance: The QueuePairNumber value is the maximum QP count.
GPU-accelerated instance: The maximum QP count is QueuePairNumber × NetworkCardQuantity.

Maximum number of QPs supported by an ERI

If you do not specify the QueuePairNumber parameter when binding an ERI, the ERI inherits the maximum number of QPs supported by the instance. This prevents binding additional ERIs even if the instance type allows more. To bind additional ERIs, change the maximum number of QPs supported by each ERI so the total does not exceed the instance limit.

Change the maximum number of QPs supported by an ERI

Call the ModifyNetworkInterfaceAttribute operation and set QueuePairNumber to the desired value.

Important

If the ERI is bound to an instance, stop the instance before changing the QP count.
For a secondary ENI bound to an instance, unbind the ENI, change QueuePairNumber, and then rebind the ENI.

View the maximum number of QPs supported by an ERI

Call the DescribeNetworkInterfaceAttribute operation and check the QueuePairNumber parameter to get the maximum QP count for the ERI.

Note

QueuePairNumber is not returned for a secondary ENI that has not been bound to an instance.

Choose an appropriate QP count

An appropriate QP count optimizes performance. Too many QPs waste resources; too few create bottlenecks.

Assess application requirements. Determine the QP count based on your application scenario, instance type, and number of ERIs. High-concurrency, low-latency applications require more QPs. Big data transfers benefit from maximizing single-QP bandwidth.
Consider hardware limits.
- The maximum QP count varies by instance type. See Maximum number of QPs supported by an ECS instance.
- Each QP consumes system resources. Factor in memory size.
Test and tune. After configuring ERIs, test performance in your environment. Adjust the QP count based on results. Watch for error messages or increased packet loss.
Monitor continuously. Regularly check the system status to ensure configurations remain suitable as workloads change.

Plan and adjust the ERI configurations based on your workload, infrastructure, and performance requirements.