Virtual IDC on the cloud

Why does cloud based business require resource certainty and service continuity

Cloud computing is evolving towards infrastructure similar to hydropower and coal, supporting users to use it on demand and pay as needed. At present, various cloud service providers at home and abroad are collaborating with ecological partners to improve the rapid iteration and promotion of cloud products and services. However, the reality is very skinny: users still face the dilemma of occasional failure to purchase specific computing product instances on the cloud in specific availability zones. The computing concept of cloud services - anytime, anywhere elasticity, why doesn't work in this scenario? Let's analyze and analyze.

At present, throughout the entire lifecycle of customer cloud business, there is a need for a "commodified" carrier that perceives computing power. For example, when a customer A migrates their personal blog web service to Alibaba Cloud, they need to purchase an Alibaba Cloud elastic computing cloud server. Customers need to be aware of cloud server specification information, such as the latest ECS. g7. xlarge. For example, customer B deploys their online 3D creative effects business on Alibaba Cloud, relying on Alibaba Cloud's powerful GPU and other computing resources. In this case, they need to purchase an Alibaba Cloud elastic computing GPU cloud server, such as ecs. gn7i c8g1.2xlarge.

Popular understanding: Similar to users renting a "room" from a "hotel". On cloud environment, users purchase a specific computing instance specification on the cloud.

This is different from the "plug and play" of hydropower and coal: the computing power on the cloud needs to perceive commodity instance information. Hydropower coal is a unified 'usage' that shields backend suppliers (which power grid supplies and which transmission line delivers) and production equipment supplied (hydroelectric power, thermal power, wind power, solar power, etc.). At present, the mainstream of computing power service sales entities for the top cloud service providers both domestically and internationally is still computing power corresponding to specific products. Due to targeting specific products, there are differences in service characteristics, suitable business scenarios, and required quantities between products. Cloud service providers also need to prepare different products and supply quantities in advance in different regions.

Because it is difficult to accurately predict the user level, purchase time, and purchase quantity of various specific computing power products, once industry hotspots arise, most customers in the same industry purchase products with a certain characteristic in large quantities in a short period of time, which is more likely to lead to specific product panic buying and some users' purchase failure. Typically, in the context of the epidemic, the rise of mining and online education has led to a strong demand for local disk and video encoding and decoding computing power, leading to a prominent phenomenon of flash buying of related products.

Popular understanding: The remaining rooms similar to "hotels" have been used up, and new customers have failed to check in. Corresponding to the cloud environment, if users purchase computing instances on the cloud and sell out short-term inventory, the purchase may fail.

Popular understanding: For expected events such as the Olympics, when users stay in a hotel, the insurance measure is to book a room in advance. For cloud environments, booking a virtual IDC (private pool) on the cloud allows for deterministic delivery of resources on the private pool.

Figure 1- Comparison of the current "service forms" between hydropower and coal infrastructure and cloud computing infrastructure

Based on the above analysis, at present, in the context of the mainstream sales form of cloud services still being "computing power commodification", users need to perceive the product characteristics required by the business's lifecycle process on the cloud, and cloud platforms need to supply and produce goods. Due to changes in demand and uncertainty in the market environment, short-term mismatches between supply and demand are more likely to occur. Therefore, serving specific customers in specific industries and making deterministic purchases for specific computing power products, namely cloud based resource deterministic delivery, has become an important ability to solve this dilemma.

How to ensure the certainty of business resources and service continuity on the cloud

The objective current situation was analyzed earlier, and there is a phenomenon of short-term purchase failure for specific regions, specific time periods, and specific computing power products. For customers, it is necessary to combine their own scenarios, the supply of cloud goods in the market, and appropriate cost inputs to achieve certainty in resource delivery, in order to ensure business continuity.

The following analysis focuses on the overall concept, and specific case studies are needed to analyze the customer's business scenario. For example, the selection of reserved regions, instance specifications, reserved duration, reserved quantity, and the optimal total cost. A division of resource delivery is shown in Figure 2, where private pooling is an important implementation method for deterministic delivery. Based on business scenarios, we recommend the best private pool purchasing solution. This article will not introduce it for the time being, but will provide specific documentation to help users better rely on cloud products and services, achieve deterministic delivery of resources, and ensure the continuity of business services.

Alibaba Cloud Private Pool Selection and Value

1- Related concepts

Private Pool: When a user purchases products such as "Elastic Guarantee" or "Capacity Reservation" on the "Resource Guarantee" service tab of the ECS console, they obtain a deterministic inventory resource reservation on the cloud, which is exclusively allocated for use. As shown in Figure 5- Private Pool Pattern Abstract and Multiple Product Implementations. On the left side of Figure 5, a private pool service has two stages: private pool reservation and private pool resource delivery. For private pool reservation, the product goal is to fulfill: ensure that the private pool is truly used. For example, EA elastic assurance requires a one-time pre charge for this private pool fee.

ICR: Immediate Capacity Reserve takes effect immediately and reserves CR as needed. The private pool is fully used up without any additional cost. Only when the private pool has remaining capacity, a partial fee will be charged for the remaining capacity.

ACR: Advance Capacity Reservation refers to a capacity reservation that takes effect at a specified time and is delayed. A deposit is charged based on credit rating. The higher the credit rating, the lower the deposit.

For the delivery of private pool resources, the product goal is: deterministic delivery and zero threshold use. After the instance is opened, it will be charged normally based on the instance.

Resource guarantee: Resource guarantee is a full chain resource deterministic service that includes quantitative perception of resource supply, deterministic reservation of resources, and planning and utilization of private pools. It can comprehensively enhance your experience in querying, booking, purchasing, and using resources, allowing you to still enjoy proprietary guarantee resources in complex and ever-changing market environments.

Elastic protection: Through elastic protection, you only need to pay a lower protection fee in exchange for a fixed period (supporting 1 month to 5 years) of resource certainty protection. When purchasing elastic protection, set attributes such as availability zone, instance specifications, and protection quantity. The system will reserve a specified number of resources that match the attributes in a private pool, such as reserving 10 ECS. c6. large instances in availability zone I of East China 1 (Hangzhou). During the validity period of elastic protection, you can enjoy resource certainty protection by choosing to use the capacity of a private pool when creating a pay as you go instance. During the validity period of elastic protection, you can repeatedly create/release a specified number of instances without worrying about resource supply issues. When the validity period of elastic guarantee is exceeded or there is no available capacity for elastic guarantee, resource deterministic guarantee will no longer be provided.

Immediate Effective Capacity Reservation: You can purchase an immediate effective capacity reservation at any time. Once the reservation is successful, it takes effect immediately and you can enjoy resource certainty services. After the capacity reservation takes effect, it will be charged based on the instance rate by volume until the capacity reservation expires and is automatically released or manually released in advance. When purchasing immediately effective capacity reservation, set attributes such as availability zone, instance specification, operating system type, capacity size, etc. The system will reserve a specified number of resources that match the attributes in a private pool. During the validity period of the capacity reservation, you can enjoy resource certainty protection by choosing to use the capacity of a private pool when creating a pay as you go instance. ECS purchased through ordinary scenarios may not be able to meet your customized needs at all times due to the ever-changing supply of resources online; During the validity period of the capacity reservation, you can repeatedly create/release a specified number of instances without worrying about resource supply issues. When the capacity reservation is not in effect or there is no idle capacity available, resource certainty guarantee will no longer be provided. During the capacity reservation billing cycle, if you purchase a volume based instance and use resource determinism, the calculated resource fees for this portion of the volume based instance will be offset by some or all of the capacity reservation fees that match the volume based instance.

When a on-demand instance matches both elastic guarantee and capacity reservation, the system will prioritize selecting the private pool corresponding to the capacity reservation product for matching.

2- Private Pool Value

Value 1: Deterministic Resource Delivery

With the widespread popularity of cloud native concepts and practices, cloud based computing research and development has become the New Normal. After the customer business cloud is native, in the rapid development process of the business, there is often a demand for deterministic delivery of resources for specific scenarios, with the expectation of 100% ensuring that the business is launched, operated, and promoted according to the established plan.

Resource assurance related products provide full chain deterministic delivery capabilities.

The ability to deliver with certainty avoids the uncertainty risk of low purchase success rate caused by the rush to purchase scarce resources in a certain availability zone on the cloud from a business perspective, such as GPU large-scale instances. On the basis of elastic delivery in the existing shared resource pool, combined with deterministic delivery, it can further ensure 100% resource protection for high priority businesses. For example, we previously purchased 20 instances of A specification in bulk, and these instances will undergo business operations such as maintenance and changes. By purchasing a private pool consisting of 20 instances of A specification, we ensure that the resources in these instances have 100% certainty during the operation and maintenance process and will not be preempted by other customers. Under normal circumstances, the capacity of 20 A specification private pools is fully utilized by 20 A specification instances, without any idle capacity, thus eliminating any additional cost investment. When the number of A instances with deterministic resource usage is less than 20, such as using only 18 instances to generate 2 idle capacity, the idle capacity will be charged in seconds and billed hourly.

Value 2: Exclusive scheduling and allocation of resources

In the iterative upgrade process of deeply integrating cloud product services into customer business architecture and business evolution, in addition to deterministic delivery of resources, flexible delivery of resources has also become an important demand. Alibaba Cloud Resource Assurance Service currently supports exclusive scheduling and allocation based on cloud private pools, and there are currently two practical methods for user exclusive scheduling.

Method 1: Users schedule and allocate instances based on the matching rules of Open, Target, and None

When creating a private pool, users specify the matching attributes for the private pool: Open, Target. When creating an instance, specify the instance matching attribute Open or Target (using the Target mode requires displaying the specified private pool ID), and the backend performs attribute matching scheduling.

When the instance matching attribute value is Open, the system will prioritize creating instances from the user's private pool; If there is no matching private pool, create instances according to the shared pool process while retaining resource deterministic features. Once idle capacity is found, the system will automatically match and associate these instances with the idle private pool on time; When the instance matching property value is Target, a specific private pool is explicitly specified, and the system performs matching verification of capacity and private pool resource rules on the specified private pool. For example, private pool region, zone, instanceType, platform, payType, and other validations.

During the operation process, when the matching properties of an instance are modified, the system will perform a timely rematch between the instance and the private pool, ensuring that the instance is as closely related as possible to the private pool, thereby reducing user cost (the idle capacity of the private pool is used in a timely manner); When a private pool with a matching mode of Open is released, the system will timely rematch the instances associated with the private pool and using the Open matching mode, ensuring that the instances are as closely related to the private pool as possible, thereby reducing user costs (the idle capacity of the private pool is used up in a timely manner).

Method 2: Users schedule and allocate instances based on Tag matching rules

When users create a private pool, they specify the tag information of the private pool, and then when creating an instance, they specify the tag information. The backend can fine tune resource scheduling and allocation from the private pool or shared pool according to the customer's specified tag matching rules.

In order to lower the user usage threshold or zero threshold, regardless of Method 1 or Method 2, Alibaba Cloud Resource Assurance Service supports users to directly use Method 1 or Method 2 for resource exclusive scheduling on the basis of existing Create Instance and Run Instances interfaces. For example, after a user applies for a whitelist, the backend specifies the matching attributes when the user creates an instance as the default values based on their needs, so that the user's existing integration interface parameters do not need to be changed.