Super Node Overview
Super nodes are Tencent Cloud's newly upgraded node product form, providing users with availability zone-level, customizable specification node capabilities. Using a super node is similar to using an extra-large CVM, making resource management and scaling simpler. Super nodes support both annual/monthly subscription and pay-as-you-go billing modes.
In the annual/monthly subscription mode, users can customize node specifications and purchase the total customized node computing power on a prepaid basis. The annual/monthly subscription mode restricts user-scheduled Pods to have a scale between 1C and 8C, with a CPU-to-memory ratio less than 1:4. This mode is suitable for fixed computing power and always-on businesses. Compared to regular nodes, the per-core price of super nodes in the annual/monthly subscription mode is lower, allowing users to migrate all eligible Pods to super nodes to reduce the per-core cost of fixed resources.
In pay-as-you-go mode, you don't need to specify node specifications, and node resources are elastic and billed by CPU and memory usage. This mode applies to elastic computing power scenarios, where pay-as-you-go super nodes can be added to support resource elasticity during peak hours, so as to further reduce cluster resource costs.
Pods deployed on super nodes have the same security isolation as cloud servers and the same network isolation as Pods deployed on existing regular nodes in the cluster. However, since super nodes do not have separate public IP addresses, access to the internet requires using NAT or binding EIP mode.
Strengths
Super nodes have the following advantages over general nodes:
Reduced costs
Purchasing an annual/monthly subscription super node offers a lower per-core price compared to purchasing a regular node.
CPU and memory pricing is lower. For details, please refer to Super Node Pricing.
The annual/monthly subscription super nodes have a larger specification limit, allowing for a wider range of schedulable Pods and avoiding the waste of boundary resources when purchasing regular nodes with an annual/monthly subscription.
Pay-as-you-go super nodes feature elasticity within seconds and on-demand use, making them quite competitive for cost reduction in elastic business scenarios.
Pay-as-you-go super nodes offer true on-demand usage, reducing resource fragmentation and improving overall cluster resource utilization. By minimizing resource buffer, costs are lowered.
The billable usage of elastic resources is shortened to save costs. As pay-as-you-go super nodes are scaled out within seconds and scaled in instantaneously, scaling costs are greatly reduced.
Easy Management
Managing resources becomes easier as you only need to manage one node per availability zone instead of multiple nodes. Annual/monthly subscription super nodes support on-demand upgrades and partial downgrades, making it as simple as managing an oversized CVM node for scaling up or down. Elastic resources can be achieved by adding pay-as-you-go super nodes, providing faster and more efficient elasticity compared to regular node pools and auto-scaling groups. This not only better meets elastic demands but also saves costs for users.
Node management comparison:
Scenario | General Node | Super Node |
Purchase | Node types, specifications, and quantities need to be planned and purchased multiple times. | Only the total resource specifications in an AZ need to be collected, and only one super node with custom specifications needs to be purchased. |
Resource scale-out | Nodes need to be added or node specifications need to be batch adjusted. | Long-term expansion: Upgrade the super node configuration and adjust the total specification size. Short-term scale-out: Use pay-as-you-go super nodes, with the elastic portion scheduled to pay-as-you-go super nodes by default. |
Resource scale-in | Nodes need to be returned, and losses need to be borne. | Long-term scaling down: Downgrade the super node configuration by adjusting the total specification size (downgrade is supported once per month), and up to 3 super nodes can be downscaled. Short-term scaling down: No action is required when using pay-as-you-go super nodes. |
Manage Node | Multiple nodes need to be managed. | Only one node needs to be managed in an AZ. |
More efficient elasticity
Compared with node pools and scaling groups, pay-as-you-go super nodes simplify server purchase, initialization, and returning during scaling processes. This greatly accelerates elasticity and minimizes possible scaling failures.
Pay-as-you-go super nodes shorten the general scaling process of 4–6 minutes to seconds, greatly increasing the efficiency.
Pay-as-you-go super nodes avoid the CA, cordoning, and Pod draining processes, truly implementing lossless and instantaneous scale-in.
Billing
Super nodes support both pay-as-you-go and annual/monthly subscription billing modes.
Instance Billing Method | Monthly Subscription | Pay-as-You-Go |
Payment Method | Prepaid | Deposit upon purchase, billed hourly |
Billing Unit | CNY/month | USD/second |
Unit Price | Lower unit price | High |
Minimum Usage Duration | Minimum usage of one month | Billed by the second, settled hourly, available for purchase and release at any time. |
Configuration adjustment | Supports Configuration Adjustment | There are no specification limits |
Use Cases | Suitable for mature online businesses with long-term stable computing power requirements. | It is suitable for scenarios where the demand for computing power fluctuates greatly |
Note:
When adding an annual/monthly subscription super node, the total price will be calculated based on the total node specifications, CPU and memory unit price, and the purchase duration for prepaid payment.
Adding pay-as-you-go super nodes is free of charge, and fees will be charged after Pods are scheduled to super nodes. The backend will calculate the fees based on the CPU, GPU, and memory resources requested by workloads and the execution duration of workloads. You don't need to pay any fees in advance.
Region and Availability Zone
Users can utilize annual/monthly subscription super nodes in the following availability zones:
China
Regions | Availability Zones |
South China (Guangzhou) ap-guangzhou | Guangzhou Zone 3 ap-guangzhou-3 |
| Guangzhou Zone 4 ap-guangzhou-4 |
| Guangzhou Zone 6 ap-guangzhou-6 |
| Guangzhou Zone 7 ap-guangzhou-7 |
East China (Shanghai) ap-shanghai | Shanghai Zone 2 ap-shanghai-2 |
| Shanghai Zone 3 ap-shanghai-3 |
| Shanghai Zone 4 ap-shanghai-4 |
| Shanghai Zone 5 ap-shanghai-5 |
East China (Nanjing) ap-nanjing | Nanjing Zone 1 ap-nanjing-1 |
| Nanjing Zone 2 ap-nanjing-2 |
| Nanjing Zone 3 ap-nanjing-3 |
North China (Beijing) ap-beijing | Beijing Zone 3 ap-beijing-3 |
| Beijing Zone 4 ap-beijing-4 |
| Beijing Zone 5 ap-beijing-5 |
| Beijing Zone 6 ap-beijing-6 |
| Beijing Zone 7 ap-beijing-7 |
Kubernetes Version
Pay-as-you-go super nodes are supported by clusters on v1.16 or later.
Annual/monthly subscription super nodes currently support clusters with version 1.18.4 and later.
Pods Schedulable to Super Nodes
Clusters with super nodes support scheduling Pods with different specifications based on different billing modes.
Annual/Monthly Subscription Super Node
Supports scheduling 1C to 8C standard specification Pods (if non-standard, they will be automatically converted to standard specifications). For specifications, please refer to Super Node Schedulable Pod Description.
Supports scheduling Pods with a CPU-to-memory ratio less than 1:4.
Pay-as-you-go super node
0.25C–16C Pods are supported.
Pods with a CPU to memory ratio of up to 1:8 are supported.
GPU Pods are supported.
Scheduling to Super Nodes
Annual/monthly subscribed super nodes and annual/monthly subscribed TKE regular nodes have equal scheduling priority.
In TKE clusters with pay-as-you-go super nodes added, when annual/monthly subscribed node resources are insufficient, Pods will be automatically scheduled to pay-as-you-go super nodes. If node resources are sufficient, Pods on pay-as-you-go super nodes will be prioritized for scaling down. Additionally, manual scheduling of Pods to super nodes is also supported.
Scenarios
Super nodes are suitable for all business scenarios. The annual/monthly subscription mode is applicable to all always-on businesses with Pod specifications ranging from 1C to 8C, while the pay-as-you-go mode is suitable for elastic scenarios.
Using Super Nodes for Always-On Businesses
Advantages: Cost-effective, easy to manage
Compared to regular nodes with annual/monthly subscriptions, super nodes offer approximately 20% lower per-core prices. For always-on services with fixed computing power requirements, if most of the current Pod specifications are below 8C, you can use the annual/monthly subscription mode for super nodes. Purchase super nodes based on the total Pod specifications as needed, schedule all 1C~8C Pods to super nodes, and reduce the per-core price of cluster resources to achieve cost reduction goals.
Pay-as-you-go super nodes for elastic businesses
Strengths: Low costs and high elasticity
In elastic business scenarios, pay-as-you-go super nodes allow for scale-out within seconds to accommodate traffic surges and require less resource buffer to reduce costs.
High elasticity: scale out in seconds, and deal with burst traffic easily. It will automatically terminate the Pods when service traffic drops. Scale in with non-stop service.
Low costs: you can reduce the reserved cluster buffer, make the usage and reservation of cluster node resources more reasonable, improve the resource utilization, and reduce costs.