Alibaba Cloud's resources cover products and services. Although Alibaba Cloud's resources can be used flexibly like water and electricity, the cost implications should not be underestimated when resources are heavily used. Similar to water and electricity, different resources have different pricing dimensions and gradients. Understanding the pricing principles and planning properly can save costs. At the same time, data center resources also have heat and scarcity. Enterprise customers need to plan the supply of resources properly. Resource planning should be based on business and cost requirements, and strike a balance between immediate supply and pre-provisioning to coordinate business growth, resource failure, high availability, and pre-provisioning time.
Resource Planning Based on Business Growth Plan
Foreseeable business evolution plans can help enterprises plan and reserve resources and reflect them in cost estimation. Business evolution on the customer side involves the following:
Deploying new businesses on the cloud requires resource planning for new businesses and systems.
Growth in existing businesses requires new computing power.
Migration of existing IDC systems to the cloud.
Cloud network transformation in hybrid cloud scenarios involves the addition of new boundary devices and security products.
Increase in data storage and data transmission bandwidth.
Before going to the cloud or migrating to the cloud, enterprises need to plan and design cloud resource capacity. By means of control, cloud cost operations can be more efficient and cost-saving. Usually, capacity evaluation can be conducted from the perspectives of permanent resources and elastic resources. Permanent resources usually support online businesses, permanent tasks, etc. The capacity planning for this part usually needs to estimate based on the water level of the actual business scenario. For scenarios with higher reliability requirements, the peak water level can be set as the permanent capacity. For scenarios with lower reliability requirements, the mean water level can be used as the permanent capacity. Elastic resources usually meet the needs of burst traffic and one-time tasks. Elastic resources can be used to meet the needs of expected business traffic increases, such as promotions and exam registrations. For unexpected burst traffic and temporary tasks, elastic resources can also be used to meet the demand. Enterprise customers can conduct research on the business plans of business departments in the near future, sort out and list suitable resource specifications and quantities that have certain redundancy for the business. Enterprise customers can do resource planning on a monthly and quarterly basis, as well as on a fiscal year basis. They can also predict future resource usage based on the current resource consumption to assist in planning.
Planning Alibaba Cloud Resource Regions
The cost of Alibaba Cloud products and services may vary depending on the location, depending on the demand and local infrastructure costs. When choosing the location of resources for cost estimation, please refer to the regions and zones to obtain a complete list of Alibaba Cloud regions and zones. It is also necessary to understand the characteristics of Alibaba Cloud infrastructure and product availability in different regions, mainly considering the following aspects:
Cloud product coverage in different regions.
Price information for using Alibaba Cloud products on the product details page or the product purchase page.
Deployment plans for cloud products in various regions.
Alibaba Cloud Computing Resource Supply Planning
Instances are the smallest units that can provide computing services for enterprise businesses. Different instance types can provide different computing capabilities. Alibaba Cloud's instance families introduces all the ECS instance families available for purchase, including the characteristics, available specifications, and applicable scenarios of each instance family. The Resource Planning tool in the Resource Management section of Alibaba Cloud Resource Management provides planning capabilities for the cloud. Users can use this tool to view whether the resource supply is healthy, combine sufficiency, replenishment capability, and heat to reflect the real-time supply health of the specified specification, and make good resource procurement plans. View the provision health scores of resources, and reserve relevant resources.
Resource supply planning refers to planning resource reservations by viewing the supply of cloud resources and implementing resource reservations.
Resource specification recommendation recommends the most suitable computing resource specifications for enterprises and the resource scale that meets the computing power requirements. From the perspective of best practices, there are two ways:
Recommendation of ECS resources based on physical machine specifications. This method is suitable for scenarios where offline IDC servers need to be moved to the cloud. It can help enterprises plan capacity based on the configuration information of the offline IDC servers (including CPU model, number of CPU cores, memory size, etc.), serving as a reference for subsequent capacity planning.
Recommendation of ECS resources based on total computing power: Recommend ECS resource solutions based on Alibaba Cloud's ECS resources. This method is suitable for customers to plan the computing power of additional businesses based on the usage of existing businesses.
The resource plan can show cost estimation, helping enterprises with financial planning and resource provisioning and management proposals. Alibaba Cloud provides Best practices for selecting computing instance families based on typical application scenarios and typical applications.
Resource Billing Method Planning
Selecting the appropriate payment method based on business characteristics is the most direct way to optimize costs from the payment perspective. From the payment perspective, stable business resource requirements that are used for a long term under various business scenarios can be supported by selecting cost-saving plans and reserved instances, etc. Low configuration resources can be used initially, and then the configuration can be upgraded after observing and evaluating the resource load. Alternatively, the configuration of low-utilization resources can be reduced or released. Stateful resource requirements whose business usage cycle is clear can be supported by selecting more flexible pay-as-you-go resources, such as temporary testing and elastic scaling. Stateless and fault-tolerant businesses can be supported by cost-effective preemptible instances. Enterprises can use a combination of various resource types to achieve cost savings. For example, combining cost-saving plans with pay-as-you-go usage can achieve higher cost savings compared to using pay-as-you-go alone in scenarios where regular business and elastic computing power are required. For example, if the business scenario is that the peak class hours are from 5:00 PM to 11:00 PM, and a large number of resources need to be scaled up temporarily, pay-as-you-go instances can be used to meet the demand. For the non-peak period, pay-as-you-go instances and cost-saving plans can be used, and during the peak period, elastic scaling and scheduled scaling of pay-as-you-go instances can be used to meet the demand. For more information, see the introduction of using pay-as-you-go instances as elastic resource groups in the Auto Scaling documentation. A combination of pay-as-you-go service instances and preemptible instances can also be used to deal with stateless business scenarios with low computing power requirements. Pay-as-you-go instances can be used to meet the minimum computing power requirements. On top of that, preemptible instances can be used to greatly reduce costs. For more details, please refer to the practice documentation on elastic supply.
Quotas Planning
Public cloud resources have constraints on quotas. Viewing Entitlement Quotas allows you to see the quotas and entitlements available to your Alibaba Cloud account. When planning, you can check whether the current quotas meet the resource requirements based on the resource demand.
ECS instance quotas are divided into total quotas, reserved quotas, and guaranteed quotas based on the supply of resources. The total quota is the upper limit of the instance. The guaranteed quota is the lower limit of the guaranteed supply of the instance. Based on the usage of the instance, the system will automatically adjust the instance quota periodically to ensure that the automatic growth of the total quota can meet the daily resource consumption growth of the enterprise. If the total quota cannot meet the demand, you can apply for a quota increase yourself.
Increase the quota limits for images, cloud disks, security groups, etc.
Plan privileged projects, including real-time downgrade, image replication, image export, etc.
Elastic Assurance Plan
You can use the Elastic Assurance Plan to guarantee resources for periodic, short-term, and business-impacting resources. For example, running computing tasks only during fixed periods of each day, week, or month, and each task requires a certain amount of resources. If the certainty of resource supply cannot be guaranteed, it will affect the business. However, in other periods, the demand for resources is relatively small overall, and using reserved instances and other methods will result in resource waste. In this case, an elastic assurance plan can be considered. Through the elastic assurance plan, you only need to pay a lower assurance fee to obtain resource certainty for a fixed period (1 month to 5 years). When purchasing elastic assurance, you can set zone, instance specifications, and other attributes. The system will reserve resources matching these attributes in the form of a private pool. When creating pay-as-you-go instances, you can choose to use the capacity of the private pool to provide resource certainty. For example, financial SaaS providers require a large number of resources for monthly account reconciliation, and rendering companies need to process a certain number of rendering tasks at the beginning of every week.
Tags Planning
Using tags can help with resource management: grouping and managing resources based on tags, as well as managing costs and accounting based on tags. When creating tags, you can implement tags based on the following design principles: When planning resources, use the Collective and Detailed Principles to plan tags concurrently and prioritize tag keys. All resource objects must be bound to the planned tag keys and their corresponding tag keys. The collective and detailed principles are necessary conditions for subsequent access control, cost tracking, automated O&M, and group searches based on dimensions. Use the Mutual exclusivity principle to avoid using two or more tag keys for the same attribute meaning. Use the Limited values principle to retain only core tag values. This simplifies resource management, access control, O&M, and cost allocation processes. When planning tags, you can consider the following best practices: Considering ramifications of future changes When planning tags, consider the impact of adding or reducing tag values in subsequent work, improving the flexibility of tag modifications. When modifying tags, it may cause changes in access control based on tags, automated O&M, or related billing reports. Simplified design principle Use fixed-dimension tag keys when planning tags. The simplification design principle reduces operational errors caused by too many tag keys.
Storage Resource Usage Planning
Alibaba Cloud's Storage Capacity Units (SCUs) provide two ways to manage storage resource usage: storage capacity unit packages and storage resource packages. Storage Capacity Units (SCUs) are pre-paid storage capacity resource packages that can be used to deduct pay-as-you-go bills for various types of cloud storage products. Compared with traffic packages for pay-as-you-go ECS instances purchased together with long-term ECS instances, SCUs can be used with multiple cloud products, providing both cost-effectiveness and flexibility in resource usage. Storage resource packages help enterprises enjoy the same amount of resources at a more favorable price, thus reducing costs. You can plan the resource package type and usage using the following methods: Recommendation 1: Select the resource package type based on the usage scenario. Recommendation 2: Select the resource package type based on the billing items on the bills. Recommendation 3: Select the resource package specification based on the usage on the bills. Recommendation 4: Troubleshooting procedures for instances that are still charged by the hour after purchase. You can see the usage and remaining amount of storage capacity unit packages and storage resource packages in the Resource Instance Management Tool.
Traffic Resource Usage Planning
Plan the usage of public network traffic and traffic resources, and purchase traffic resource packages to deduct the billing items for outbound public network traffic. When designing the architecture, manage inbound and outbound public network traffic through NAT Gateway, and share the public network traffic bandwidth.
Choose the Appropriate ECS Type when Building Container Clusters
Proper planning of cluster host specifications for dedicated container clusters is recommended. It is suggested to use large ECS instances: Advantages of using large ECS instances:
Higher network bandwidth, higher resource utilization for applications requiring large bandwidth.
The proportion of containers communicating within a single ECS instance increases, reducing network transmission.
Higher efficiency in pulling images. After an image is pulled, it can be used by multiple containers. If small ECS instances are used, the image needs to be pulled multiple times. If ECS cluster scaling is required, more time will be spent. Immediate response cannot be achieved.
When creating Kubernetes clusters, many smaller ECS instances are usually used. This has the following disadvantages:
Limited network resources of small ECS instances.
If the occupancy rate of a small ECS instance is close to 100% (each ECS instance can accommodate only one container), the remaining resources cannot be used for new containers or to recover failed containers. If there are many small ECS instances, resource waste will occur.
Resource Planning under Cost Optimization Goals
Identify and manage the cost optimization goal requirements and plan resource reduction with the following methods:
Metering and pricing vary by product, and usually have different pricing levels based on the location, size, or capacity of resources. Costs related to infrastructure are relatively small.
Extract shared resources into shared resource groups to maximize resource utilization and allocate costs to different business units.
In cost consolidation and analysis, find resources with high costs and low resource utilization for resource reduction planning.
Break down cost optimization goals into resource groups and cost units, and set cost optimization goals specifically.
Design based on best practices for cost optimization to reduce resource scale and costs through architectural optimization.
Cost optimization should be placed first. Resource planning should come before resource provisioning and management to help control costs.
Use Alibaba Cloud's saving plans.
Combine the cost optimization phase recommendations to achieve the cost optimization goals.
