Introduction to business on the cloud is normalized, and the purchase of resources on the cloud, flexible delivery, and self-service have become a major trend. Different customers in different industries have different business development stages, and the cost of resources on the cloud accounts for different proportions of the overall business cost. Minimizing cost investment and maximizing business benefits are always the common goals of different customers. Alibaba Cloud provides users in the entire industry with a wealth of cloud computing power product services and flexible and diverse sales models, helping users to accurately estimate resource capacity on the cloud and delicate resource delivery management, which is very helpful for customers to save resources purchased on the cloud. the cost of. This article is the second in a series of private pools on the cloud. It will focus on the selection guide for private pools in different scenarios.
Author of this article: Alibaba Cloud technical expert Li Yuqian
Introduction:
The normalization of business on the cloud, the purchase of resources on the cloud, flexible delivery, and self-service have become a major trend. Different customers in different industries have different business development stages, and the cost of resources on the cloud accounts for different proportions of the overall business cost. Minimizing cost investment , Maximizing business revenue is always a common goal among different customers.
Cloud provides 16189e43d357c1 rich cloud computing power product services and for users across the industry, 16189e43d357d6 flexible and diverse sales models, to help users accurately estimate the resource capacity on the cloud and fine- The cost of purchasing resources on the cloud.
This article is best practice - the second in a series of private pools cloud, in first post , the author focuses on the value and how to get private pool; This paper focuses on the optional private pools under different scenarios guide.
Let’s review what a private pool is: When a user purchases products such as “flexible guarantee” or “capacity reservation” on the ECS console, under the “Resource Guarantee” service tab, they will obtain a deterministic computing resource on the cloud (CPU And Memory) reserved, and is a resource pool exclusively allocated and used. The service of a private pool has two stages: private pool reservation and private pool resource delivery.
The private pool has the resource inventory certainty , resource scheduling and delivery flexibility , which can escort the certainty and continuous development of customer business. Then, for different customers, choosing the most suitable private pool can match the resource cost and business development.
We know that customers on the cloud come from all walks of life, realizing the industry's own value through industry digital solutions and digital products and services, and relying on the cloud platform to provide various computing services. Computing power services will eventually be reflected in changes in resource requirements. We will abstract resource requirements change characteristics shown in FIG 1, is divided into daily stability requirements , daily elastic demand , burst demand three types.
Figure 1-Characteristics of resource requirements
As shown in Figure 1, the deterministic demand for resources is concentratedly reflected in "daily elastic demand" and "bursting demand". Among them, "daily elasticity" demand can be subdivided into "cyclical" short-term resource demand and "non-cyclical" short-term resource demand (incidental and special period). In summary, the scenarios that require deterministic delivery are concentrated in:
- "Cyclical" short-term resource requirements
- "Incidental" large resource requirements
- "Special period" resource needs
The following is an introduction to the deterministic resource purchase of the three scenarios.
Periodic short-term resource needs
As shown in Figure 2, resource demand shows obvious periodicity and regularity. The change characteristics of the number of instances over time are more in line with the resource requirements of games and online education scenarios. For example, after get off work hours and weekends, the number of instances increases, but the number of instances is usually small.
Game scenario: XX games open new servers at a fixed time every Saturday, a large number of users flood in and register, and resource demand surges; online education scenario: XX online education company, the summer school schedule for online teaching is concentrated at a fixed time, and a large number of courses are generated Resource demands, the resources can be released after the course is over.
Figure 2-Periodic short-term resource demand
Deterministic delivery plan
In view of periodic short-term resource requirements, the following is a classified introduction from the length of resource instance holding time and the multi-cloud platform. Under each category, multiple purchase options are subdivided and related advantages and disadvantages are displayed.
Scheme 1
Long-term holding
The core of this plan is to purchase the resources that are periodically needed in advance at one time. As shown in Table 1:
Table 1-Long-term holding
Scheme 2
Short-term holdings
The core of this solution is to ensure the deterministic delivery of resources only when resources are needed during the cycle time period. As shown in table 2:
Table 2-Short-term holdings
In view of the cyclical short-term resource demand, the purchase of "flexible guarantee" requires a certain fee in advance. How does the cost input compare to other purchase methods? Let's do further analysis below.
Example : Suppose the user has a certain resource requirement: Beijing region, instance size ecs.g6.xlarge, 1 unit, and the estimated cumulative usage time within one month is 12 days, and other time resources can be released within one month. Business requirements: When resources are needed at any time, they must be delivered 100% successfully. At this time, to ensure the deterministic delivery of resources, users have 4 purchase and corresponding billing methods, as shown in Table 3 below:
Table 3-4 Comparison of costs
Method 1 : One month of "monthly subscription", total cost 1 = monthly subscription price of this instance * 1
Method 2 : One month of "Pack Year",
Total cost 2 = Average monthly price of the instance package year * Duration (months in this case 1)
Method 3 : 0 prepaid RI reserved for one month,
total cost 3 = one month RI cost of this instance
Method 4 : Flexible guarantee to reserve one month, total cost 4 = Guarantee package reservation cost + actual duration of the instance start time (40% * 30 = 12 days, only 40% of the time in this case per month Use); if you buy directly on a monthly basis, you need to pay for one month.
Note: The comparison of the cost of a certain instance of the specified specifications is as follows:
Annual monthly average price <monthly monthly price <monthly price accumulative by volume
Thus, in same configuration conditions the above four different ways cost relationships are optional: 1 Total Cost> Total Cost 2> The total cost of 3> Total Cost 4
This time, in same configuration under the "flexible protection +12 days," according to the total cost of the optimal amount of hours open.
The relationship between holding time and cost in the four ways is abstracted as shown in Figure 3:
Figure 3-Comparison of cost and duration of various methods
Note : The price here is a schematic diagram, because of the different configurations of the instance, the final price is different, and the price of the instance will also be adjusted due to market changes. Diagram 3 above focuses on expressing "comparative information": when the use time is less than 40%, the flexibility guarantee is relatively cost-saving. When the usage time exceeds 40%, the flexibility guarantee is no longer cost-effective compared to a monthly subscription.
Scheme 3
* * Mixed period holding
The features of this program: business fine management of resources . For example, big data calculation processing, fixed purchase of instances every night, or a large number of purchases of instances during a big promotion event. In these scenarios, computing tasks show "level characteristics": the main tasks have strict delivery time constraints (resources need to be guaranteed first), Delivery of non-primary tasks can be delayed (depending on the resources of Spot auction instances for processing). The optional purchase plans are shown in Table 4:
Table 4-Mixed cycle holdings
Scheme 4
cloud combination
The feature of this scheme: When a cloud platform temporarily cuts the supply of a specific product, it will be transferred to another cloud platform for instance purchase.
Table 5-Multi-cloud combination
Incidentally large demand for resources
Occasionally a large number of resource requirements show unpredictability and suddenness, as shown in Figure 4. For example, Internet or To C service customers, some hot events trigger a large number of service requests, which temporarily require a lot of resources to deal with.
Figure 4-Incidentally large demand for resources
Deterministic delivery plan
Since it is an occasional large resource demand, a feasible solution: insurance service . Alibaba Cloud's elastic guarantee is the product and service spawned under the background of the 2019 new crown epidemic. The user only needs to pre-deliver a certain fee (the cost is related to the amount of resource demand, the guarantee time, and the number of openings), and then according to the agreement, the user can obtain deterministic resource delivery when there is a sudden resource demand. Related purchase plans are shown in Table 6.
Table 6-Incidental resource purchase plan
Resource needs in special periods
As shown in Figure 5, the resource demand in a special period shows planning , predictability . For example, Double 11, Spring Festival, statutory long holidays, etc., require resource protection to ensure that services are carried out in an orderly manner.
Figure 5-Resource demand in a special period
Deterministic delivery plan
For resource needs in special periods, long-term holding is generally not recommended, because short-term resource needs and long-term costs are not cost-effective. Related optional purchase plans are shown in Table 7.
Table 7-Resource Purchase Plan in Special Period
Combination recommendation
The above focuses on analyzing multiple purchase scenarios from a single perspective of "deterministic" resource delivery. Actual customer "as a whole" Resource demand (traffic load) also exhibit certain characteristics, as shown in Figure 6, we abstract resonance , smooth type , thrusting type , mixed fabric type four categories.
Figure 6-Classification of business load characteristics
For the four classification characteristics of business load, the corresponding purchase recommendation is shown in Figure 7:
Figure 7-Combination recommendation
Summarize
In response to “deterministic” resource delivery scenarios, Cloud provides a variety of options, such as 16189e43d36731 elastic protection, , immediate effective capacity reservation , delayed effective capacity reservation and other product services, especially e-commerce 618 , Double 11 and other predictable short-term resource demand scenarios, flexible guarantee or delayed effective capacity reservation is the preferred solution. For periodic resource demand scenarios, where deterministic delivery is required, capacity reservation with immediate effect combined with SavingPlan is the preferred solution. With deterministic resource delivery, the fine management of resources and other tasks have a foundation. For more knowledge about cluster resources, you can refer to the book "Deep Clusters-Resource Scheduling and Management of Large Data Centers".
This is the end of the sharing of best practices in this issue. Soon we will launch the third part of the private pool series on the cloud-how to plan and implement the capacity of my business after the cloud, so stay tuned~
Related Reading
Copyright Statement: content of this article is contributed spontaneously by Alibaba Cloud real-name registered users. The copyright belongs to the original author. The Alibaba Cloud Developer Community does not own its copyright and does not assume corresponding legal responsibilities. For specific rules, please refer to the "Alibaba Cloud Developer Community User Service Agreement" and the "Alibaba Cloud Developer Community Intellectual Property Protection Guidelines". If you find suspected plagiarism in this community, fill in the infringement complaint form to report it. Once verified, the community will immediately delete the suspected infringing content.
**粗体** _斜体_ [链接](http://example.com) `代码` - 列表 > 引用。你还可以使用@来通知其他用户。