Configure highly available network connectivity for workload public endpoints
When you build public-facing business systems, you must ensure that users can access your workload's public endpoints reliably, securely, and efficiently. This is critical to overall system availability. To do so, use highly available DNS services, a content delivery network (CDN), API Gateway, load balancing, and elastic backend services.
Priority
High
Not recommended
Unplanned high availability DNS and unused high availability public network resources
You can directly access an instance using its IP address instead of a domain name.
Do not use A records to access Application Load Balancer (ALB). Use CNAME records instead. If you use A records, traffic cannot be automatically redirected to healthy instances if the underlying IP address or its load balancer becomes unhealthy.
Do not deploy backend servers for a load balancer in a single zone. This creates a single point of failure because if that zone experiences a network outage or hardware failure, your entire service becomes unavailable. Instead, deploy backend ECS instances across multiple zones.
Do not mix high-priority and low-priority workloads in the same public network resource pool without isolating their bandwidth and IP addresses. A traffic burst from a low-priority workload can consume all resources and degrade the performance of high-priority workloads.
Implementation guide
Use a highly available DNS service. Alibaba Cloud DNS provides a highly available and scalable DNS service that ensures domain name resolution and supports health checks. Traffic is routed only to healthy ports. For more information, see What is Alibaba Cloud DNS.
Use Alibaba Cloud Global Traffic Manager (GTM) to provide a unified entry IP address and optimize global access paths. GTM supports intelligent scheduling based on factors such as geography, ISP, and latency. For more information, see What is Global Traffic Manager 3.0.
Use Alibaba Cloud CDN to accelerate static resource delivery and improve availability. Edge caching reduces the load on the origin server and improves response speed. For more information, see What is Alibaba Cloud CDN.
Split public network resources and load balancer (LB) instances:
Split public bandwidth resources based on business type (such as core business or staging environment), traffic characteristics (such as high concurrency or persistent connections), and stability requirements. Assign dedicated bandwidth and IP addresses to high-priority workloads, such as payments or core transactions. This prevents low-priority traffic, such as log synchronization or backend management, from consuming excessive bandwidth.
Split LB instances. Route traffic from different workloads or customer groups to separate LB instances to avoid global impact from a single instance failure.
Configure CNAME-based access for LB instances:
Use a custom domain name. Point the custom domain name to the domain name of your Network Load Balancer (NLB) or Application Load Balancer (ALB) instance using a CNAME record. For more information, see Configure CNAME records for ALB.
If the service associated with your CNAME record fails, you can update the authoritative DNS CNAME record to an A record that points to the Elastic IP Address (EIP) of the NLB or ALB instance for failback.
If a public IP address of an NLB or ALB instance becomes unavailable, you can request a new EIP to replace it.
Deploy ingress-layer resources across multiple zones:
Create a multi-zone Server Load Balancer (SLB) instance.
Create a backend server group and ensure that the servers in the group span multiple zones.
Configure listeners, forwarding rules, and health checks. For more information, see Build an application delivery network with SLB.