High-performance Classic Load Balancer (CLB) instances provide reliable performance metrics. This topics provides answers to some frequently asked questions about high-performance CLB instances.

What is a high-performance CLB instance?

High-performance CLB instances provide reliable performance metrics. Shared-resource CLB instances share resources with each other and their performance is not guaranteed.

Before Alibaba Cloud launched high-performance CLB instances, all CLB instances were shared-resource instances. You can view the instance type in the CLB console.
Important Shared-resource instances are no longer available for purchase.

The following table describes the differences between shared-resource CLB instances and high-performance CLB instances.

Item Shared-resource CLB instance High-performance CLB instance
Resource allocation Shared resources Exclusive resources
Service uptime guaranteed by terms of service level agreement Not supported 99.95%
IPv6 ×
Server Name Indication (SNI) ×
Blacklists and whitelists ×
Associating with elastic network interfaces (ENIs) ×
Adding secondary IP addresses of ENIs that are associated with Elastic Compute Service (ECS) instances ×
HTTP-to-HTTPS redirection ×
Consistent hashing ×
TLS security policies ×
HTTP2 ×
Websocket(S) ×

You can move the pointer over the question mark that corresponds to the specification of a high-performance CLB instance to view the performance metrics, as shown in the following figure.

Performance metrics of high-performance CLB instances

The following section describes the metrics of high-performance CLB instances:

  • Maximum number of connections

    The maximum number of concurrent connections that a CLB instance supports. When the number of existing concurrent connections reaches the upper limit, new connection requests are dropped.

  • Connections per second (CPS)

    The number of new connections that are established per second. When the CPS value reaches the upper limit, new connection requests are dropped.

  • Queries per second (QPS)

    The number of HTTP or HTTPS queries (requests) that can be processed per second. This metric is specific to Layer 7 listeners. When the QPS value reaches the upper limit, new connection requests are dropped.

The following table describes the specifications of high-performance CLB instances provided by Alibaba Cloud. The available specifications vary based on the region. You can go to the buy page of the CLB console to view the available specifications in each region.

Table 1. Specifications of high-performance CLB instances
Specification Maximum number of connections CPS QPS
Small I (slb.s1.small) 5,000 3,000 1,000
Medium I (slb.s2.small) 50,000 5,000 5,000
Medium II (slb.s2.medium) 100,000 10,000 10,000
Large I (slb.s3.small) 200,000 20,000 20,000
Large II (slb.s3.medium) 500,000 50,000 30,000
Super Large I (slb.s3.large) 1,000,000 100,000 50,000
Note
  • The maximum capacity of a pay-by-LCU CLB instance is the same as a Super Large I (slb.s3.large) CLB instance.
  • If you require a larger number of connections at Layer 4, use Network Load Balancer (NLB). If you require more queries per second at Layer 7, use Application Load Balancer (ALB).
    • For more information about NLB, see NLB.
    • For more information about ALB, see What is ALB?.

Can I change the specification of a high-performance CLB instance?

Yes, you can change the specification of a high-performance CLB instance in the console. For more information, see Modify the configurations of pay-as-you-go CLB instances.

Note
  • You can change only shared-resource CLB instances to high-performance CLB instances. You cannot change high-performance CLB instances to shared-resource CLB instances.
  • Changing a shared-resource CLB instance to a high-performance CLB instance does not negatively affect your workloads, or change the IP address of the CLB instance.

    However, we recommend that you change a shared-resource CLB instance to a high-performance CLB instance during off-peak hours. You can also configure DNS resolution to replace your current load balancing service before you change the CLB instance.

  • The IP addresses of CLB instances are not affected when you change instance specifications.

How are the fees for high-performance CLB instances calculated?

For more information about pay-as-you-go CLB instances, see Pay-as-you-go.

Pay-by-specification

Total fees for an Internet-facing high-performance CLB instance = Instance fee + Data transfer fee or bandwidth fee + Specification fee

You are charged a specification fee on an hourly basis based on the specification that you select. For example, if you select the Super Large I (slb.s3.large) specification, but the performance metrics of an hour are the same as the performance metrics of a Large I (slb.s3.small) CLB instance, you are still charged a specification fee based on the Super Large I (slb.s3.large) specification for the hour.

Pay-by-LCU

Total fees for an Internet-facing high-performance CLB instance = Instance fee + Data transfer fee + LCU fee

If you use the pay-by-LCU metering method, you do not need to specify a specification for a CLB instance. The CLB instance automatically adds or releases resources on demand. You are charged an LCU fee based on the actual resources that are consumed per hour.

How do I choose among the specifications?

  • If your workloads expect only minor fluctuations, we recommend that you select pay-by-specification. You are charged a specification fee on an hourly basis based on the specification that you select.
  • If your workloads expect periodic or large fluctuations, we recommend that you select pay-by-LCU. Pay-by-LCU CLB instances can automatically scale resources on demand. In addition, you do not need to specify a specification. You are charged for consumed LCUs on an hourly basis.
Comparison between pay-by-LCU and pay-by-specification

Am I charged specification fees for existing shared-resource CLB instances?

No, you are not charged specification fees for existing shared-resource CLB instances.

You are charged specification fees only if you upgrade the shared-resource CLB instances to high-performance CLB instances. After the free trial for high-performance CLB instances ends, you are charged specification fees if you use high-performance CLB instances.

Important Shared-resource instances are no longer available for purchase.

Why is a high-performance CLB instance unable to reach the performance limit defined in the specification?

This issue can be explained by the cask theory.

High-performance CLB instances do not guarantee that the three metrics can reach the upper limits of a specification at the same time. If one of the metrics reaches the upper limit, the instance performance is limited.

For example, you purchase a high-performance CLB instance of the slb.s3.small specification. When the QPS value reaches 20,000, new connection requests are dropped by the CLB instance even if the number of concurrent connections has not reached 200,000.

Are shared-resource CLB instances still available for purchase?

No, shared-resource CLB instances are no longer available for purchase.

Shared-resource instances are no longer available for purchase.

Am I charged a specification fee for an internal-facing CLB instance?

You are not charged a specification fee for a shared-resource internal-facing CLB instance. You are charged a specification fee for a high-performance internal-facing CLB instance.

The specification fee for an internal-facing CLB instance is calculated in the same way as the specification fee for an Internet-facing CLB instance. You are not charged instance fees or data transfer fees when you use internal-facing CLB instances.

What do I do if I need more high-performance CLB instances?

If the highest quota that you can apply for on the Quota Management page cannot meet your business requirements, you can apply for the slb_privilege_allow_more_guaranteed_performance_instances privilege. This privilege allows you to own more high-performance CLB instances but does not allow you to own more shared-resource CLB instances. For more information, see Quota management.

Why does the instance type remain unchanged after I upgrade a shared-resource CLB instance to a high-performance CLB instance?

If you change the metering method (pay-by-bandwidth or pay-by-data-transfer) when you upgrade a shared-resource CLB instance to a high-performance CLB instance, the new instance type and the new metering method take effect at 00:00:00 the next day.