Performance testing
How do I test the packet forwarding capability of an instance?
What do I do if performance is unstable during UDP PPS or TCP bandwidth tests on an ECS instance?
High bandwidth usage
Troubleshoot and resolve high network bandwidth usage on a Linux instance
Troubleshoot and resolve high network bandwidth usage on a Windows instance
Traffic exceptions
Performance optimization
Appendix
How do I test the packet forwarding capability of an instance?
For general methods to test packet forwarding, see Network performance test methods. For the recommended method to test packet forwarding for g7, c7, and r7 instances, see Use Pktgen to test network performance.
What is the packet loss probability for communication over the internet between ECS instances in different regions?
When ECS instances in different regions communicate over the internet, the P99 packet loss rate per hour is less than 0.0001%.
What is the network latency for internal network communication between ECS instances in the same zone of the same region?
ECS instances that communicate over the internal network within the same zone of a region have the lowest latency. The P99 one-way latency for communication between instances in the same zone is less than 180 microseconds (μs).
You can view the latency between different zones.
How do I ensure connection performance for an instance that does not have a specified connection limit?
If the performance metrics for an instance family do not specify a connection limit, connection performance is not guaranteed for that instance family. You can run a stress test on your services to understand the instance performance and select a suitable instance type.
After a connection is established, it has an aging time. The connection continues to occupy a connection slot until the aging time expires. Therefore, the number of occupied connections may be greater than the number of connections that are in use by your services.
What do I do if performance is unstable during UDP PPS or TCP bandwidth tests on an ECS instance?
When you run a network performance test on an ECS instance, many factors can affect the test results. These factors include common performance tuning methods such as Non-Uniform Memory Access (NUMA) topology adaptation, binding tasks to vCPUs, and binding interrupts to vCPUs.
For example, during a single-stream TCP bandwidth test, if the receiving task (such as a netserver process) and the network interface card (NIC) receive queue interrupt are bound to the same vCPU, the NIC triggers an interrupt that disrupts the receiving task when a data frame is received. Frequent disruptions to the receiving task can cause unexpected test results. To resolve this issue, you can bind the receiving task and the NIC receive queue interrupt to different vCPUs. This leverages the performance advantages of multiple vCPUs to achieve better test results.
Does the public bandwidth of an ECS instance fluctuate?
The Pay-by-bandwidth billing method guarantees the bandwidth value that you set. The value does not fluctuate due to other factors.
For the Pay-by-traffic billing method, the peak inbound and outbound bandwidth values are upper limits and are not guaranteed. When resource contention occurs, the peak bandwidth may be limited. If your services require guaranteed bandwidth, use the Pay-by-bandwidth billing method. For more information, see Network bandwidth.
Why does a new ECS instance have about 200 Kbps of inbound traffic?
This traffic is generated by Address Resolution Protocol (ARP) broadcast packets. Each ECS instance is assigned to a large network segment. In the same network segment, your new ECS instance receives a request message whenever an ECS instance sends an ARP broadcast to the gateway. However, your ECS instance does not need to send a response message if the request is not for its IP address.
Why is the bandwidth for an ECS instance shown in Cloud Monitor different from the bandwidth shown in the ECS console?
An ECS instance acts as a backend server for a Server Load Balancer (SLB) instance and uses the Layer 7 HTTP forwarding model. In this model, the SLB instance forwards client requests to the ECS instance. The ECS instance then responds to the client over the public bandwidth. The bandwidth used by this response is not displayed in the ECS console. It is included in the outbound traffic of the SLB instance. For this reason, the bandwidth shown in Cloud Monitor is different from the bandwidth shown in the ECS console.
Why does a stopped ECS instance with a pay-by-data-transfer public bandwidth still generate outbound traffic?
Symptom: In Anti-DDoS Origin Basic, the status of the ECS instance is Cleaning, while the instance itself is Stopped. However, fees are still generated every hour for the pay-as-you-go bandwidth.
Cause: CC protection is enabled for the ECS instance. After CC protection is enabled, the security mechanism actively sends probe packets to detect the attack source. This may generate a large amount of outbound traffic.
Solution: You can disable CC protection. For more information, see Configure CC protection.
How do I disable the ECN feature on Windows?
If you find that network connectivity is unstable, the network speed is slow, or specific applications behave abnormally after you enable Explicit Congestion Notification (ECN), you can disable ECN to resolve the issue.
Log on to the ECS instance.
Go to ECS console - Instances. In the top navigation bar, select the target region and resource group.
Go to the details page of the target instance, click Connect, and select Workbench. Set the connection method to Terminal, enter the username and password, and then log on to the graphical terminal page.
In the taskbar search box, enter
PowerShelland open it.Disable the ECN property.
netsh interface tcp set global ecncapability=disabledCheck whether ECN is disabled.
netsh interface tcp show globalKey property: If the ECN Capability property is set to disabled, ECN is disabled.
What is the maximum bandwidth for an ECS instance?
ECS bandwidth includes internal bandwidth and public bandwidth.
Internal bandwidth: The internal bandwidth of an ECS instance depends on its instance type. For the exact value, refer to the basic network bandwidth metric in the Instance family list.
Public bandwidth: If you use a static public IP address, the maximum bandwidth is 200 Mbps for the pay-by-bandwidth billing method and 100 Mbps for the pay-by-data-transfer billing method. The actual limit is shown on the purchase page. For more information, see Modify bandwidth configurations. If the peak bandwidth of a static public IP address does not meet your requirements, you can use an Elastic IP Address with Internet Shared Bandwidth.