CLB guarantees high availability (HA) from multiple dimensions such as system architecture and product configuration. Based on the business scenario and requirements, you can select one from a variety of HA solutions for cross-region disaster recovery or intra-region cross-AZ disaster recovery.
CLB Cluster HA
CLB instances adopt a cluster deployment mode that supports session synchronization, eliminating the single point of failure for servers, enhancing system redundancy, and ensuring service stability. All CLB instances have cluster-level high availability.
At Layer-4, load balancing is implemented based on the unified Tencent Gateway (TGW), which features high availability, strong scalability, high performance, and strong anti-attack capability. It supports high-performance forwarding based on Data Plane Development Kit (DPDK). With TGW, a single cluster can support hundreds of millions of concurrent requests and tens of millions of packets per second (PPS). Many Tencent businesses, such as Tencent Games, Tencent Video, WeChat, and QQ, use TGW for service access.
At Layer-7, load balancing is implemented based on Secure Tencent Gateway (STGW), which is a load balancing service developed by Tencent based on Nginx. It supports large-scale concurrency and carries a large amount of Tencent's Layer-7 business traffic.
Single CLB Instance HA
Non-Domain Name-Based Public Network CLB
The non-domain name-based public network CLB provides services in the form of VIP addresses, with the SLA of 99.95%. There are two deployment schemes for associated clusters of the VIP addresses:
Deployment Mode
Cluster Disaster Recovery
Cross-AZ Disaster Recovery
Single-AZ
Supported
Not supported
Multi-AZ
Supported
Supported. The primary/secondary AZ mode is adopted. In case of a failure in the primary AZ, the CLB can automatically switch to the secondary AZ and restore services within a very short time.
Domain Name-Based Public Network CLB
The "domain name-based public network CLB" builds upon the "non-domain-name-based public network CLB" mentioned above by adding an additional DNS service layer. The SLA is increased from 99.95% to 99.99%, and it supports the replacement of faulty VIP addresses to enhance availability. For details, see Announcement on the Launch of Domain Name-Based Public Network CLB.
Private Network CLB
Private network CLB instances are deployed with the nearby access architecture, which means that one CLB instance is deployed to one or more AZs. When a client accesses such a CLB instance, the cluster with the lowest latency in these AZs is selected automatically to forward the access traffic to a backend server. If the selected CLB cluster in an AZ becomes unavailable, the system can switch to the CLB cluster in another AZ.
Multi-CLB Instance HA
If you have very high availability requirements, the availability mechanism of a CLB instance itself may not meet your needs in scenarios such as network attacks, cross-region switching, and configuration errors. You can create multiple CLB instances and use Tencent Cloud DNS to schedule the access traffic.
Comparison Between the Multi-CLB Instance HA and Domain Name-Based Public Network CLB:
Comparison Item
Domain Name-Based Public Network CLB
Multi-CLB Instance HA
SLA
99.99%
99.95%
Input Failover
Provide link detection and disaster recovery switch capabilities, eliminating concerns about a single IP address entry interruption. When a single IP address experiences a failure, the faulty IP address can be switched to reduce business impact.
Depends on your configured DNS resolution and switching policies. Prompt detection and switching is required for your business.
Ops Management
Only a single instance needs to be configured.
Multiple CLB instances and corresponding DNS resolution policies need to be configured.
Cost
The cost is lower for only CLB-related fees are charged.
Deployment of multiple CLB instances and components such as DNS resolution leads to higher costs.
Region
The associated cluster of the CLB instance is deployed in a single region.
You can select CLB instances across multiple regions.
Best Practices:
If your business is deployed within a single region, it is recommended to choose the domain name-based public network CLB solution first, which supports the switch of faulty IP addresses.
If your business is deployed across multiple regions and has very high disaster recovery requirements, it is recommended to select the multi-CLB instance HA solution.
When a client accesses the same port of a real server via different intermediate nodes at the same time, these nodes cannot be specifically determined. For details, see the issue description.
Real Server HA
CLB determine the availability of real servers through health checks to avoid real server exceptions from affecting frontend businesses, thereby improving the overall availability of businesses.
After the health check feature is enabled, CLB instances will perform health checks regardless of the real server weight (including 0). You can view the health check status in the Health Status column on the instance list page or on the details page of the real server bound to the listener. For details of the health check mechanism, see Health Check Overview.
