Introduction to ECS Purchasing Options
Introduction to ECS Purchasing Options

This course introduces Alibaba ECS purchasing options, including ECS purchasing options, preemptible instances, reserved instances, saving plans, and cost optimization.

Learning Objectives

  • Get a solid understanding of Alibaba ECS purchasing options

Intended Audience

  • Solution architects
  • System operators
  • Developers
  • Anyone who wants to learn about ECS purchasing options


To get the most out of this course, you should have a basic understanding of Alibaba Cloud ECS.


Welcome to join this training session. My name is Lucas. I'm the ECS Product Solution Architect from Alibaba Cloud. For today, I'm going to share with you guys about the introduction of ECS purchasing options. In today's session, we will go through the ECS purchasing option details, and also take a look on how customer can optimize the cost by using the available options. Without further ado, let's start with the overview of ECS purchasing options.

As you can see from the below tables, there are two major category of payment options that customer can choose when purchasing an ECS instance. Pay-as-you-go model, which based on the on demand basis and subscription model, which is a complement of long-term usage. Under the pay-as-you-go category, there are two options which are the preemptible instance and on demand instance. Both of them are having the same charging mechanism, which is per second billing and hourly charge.

Preemptible instance, we have a lowest hourly charge if compared to the rest, whereas on demand instance is based on the regular pricing which defined by the Alibaba Cloud. For subscription model, the charges are according to the subscription term that a customer define which can be weekly, monthly, or yearly. Definitely, the longer subscription period that customer subscribe, the lower hourly price rate they could have. And please kindly take note that 3 years and 5 years subscription plans are not available in international sites.

In pay-as-you-go option, customer has the flexibility to decide when to release the ECS instance. However, for subscription ECS instance, it only can be released after end of the subscription period. Both reserved instance and saving plan are the cost optimization plan or pay-as-you-go instance. Reserved instance is having a flexibility and predictability, which can provide up to three years reservations with a better discount rate. Whereas for saving plan, they provide a more flexible option than reserved instance without need to specific resource reservations.

In the next slide, we are going to discuss further on preemptible instance. Preemptible instance is a type of on-demand instance that are offered at a discounted price compared to the normal pay-as-you-go instance. The preemptible instance is protected for an initial one hour period. The resource can be reclaimed after the protection period subject to the inventory level and also the market price. But the price of a preemptible instance covers only the vCPU and memory. The rest of resources such as system disks, data disks, and network bandwidth are charged on the normal pay-as-you-go basis.

The preemptible instance are applicable for ECS and ECI, and there are two bidding box for preemptible instance. Set maximum price by customer and system will do the comparison with customer bid price and also the current market price. The next is use automatic bid. System will bid automatically by using the price of a pay-as-you-go instance as the maximum price of customer. Preemptible instance are supported for automatically deployment method such as auto scaling service, auto provision group, ACK/ASK, Kubernetes, and Terraform.

Here is a lifecycle diagram of a preemptible instance. When customer create a new preemptible instance, the system will compare if the customer bid price is higher than the current market price, and the stock of the instance that the sufficient. Once the above two criteria can be fulfilled, then the instance will be created successfully and run for an hour during that initial protection period. The system will start to check again the resource inventory and also the pricing comparisons with the latest market price for every five minutes after the production period ended.

The instance will continue to run until customer release it if it about to fulfill the condition of inventory check and also the current market price comparisons. Otherwise, the instance will be released by Alibaba Cloud. Either the customer bid price is lower than the market price or inventory is insufficient. Below are some use cases for preemptible instance. Preemptible instances are suitable for non-production workloads and stateless application, such as real-time data analysis, image and media coding, development test environment and also for other stateless business scenarios.

For those stateful applications such as database are strongly not recommended to use preemptive, for instance, due to the nature behavior. Next, we will take a look on the assemble of building for preemptible instance. As you can see from the table below, the customer purchases a preemptible instance with defined bid price of $11. And during that time, the market price is $10 at 8:40 A.M. Instance is going to be created because of the customer bid price is higher than the market price. And at the next moment, market price increased to $15 at 8:50 A.M. However, the VM will not be released due to within one hour protection period. 

At 10 o'clock, the market price increased to $15 again, which is higher than the user defined bid price. And during this time, customer will receive a notification that VM going to release in five minutes due to after one hour protection period. So the total bidding will be $10 for the first initial hour and subsequent 10 minutes with a different hourly charge for $1 and 33 cents and also the 83 cents. So the total will be $12 and 16 cents, in this case.

Next, we will move into the topic of reserved instance. Reserved instance is a discount coupon that can be automatically applied to one or more pay-as-you-go instances, excluding preemptible instances. There are two types of reserved instance, regional reserve instance, it supports different size of matching, zone flexibility. However, resource reservation is not supported for regional reserve instance. The second one, zonal reserve instance. Different size matching is not supported. Zone flexibility is not supported. However, resource reservation is supported for the zonal reserve instance. Both regional and zonal reserve instance are able to subscribe from one year and up to three years.

There are three payment models, which is no upfront, partial upfront, and all upfront and definitely for all upfront method, we will provide better discount rate to the customer. Below are the reserve instance attributes. So as long as the on demand instance can match with the following attributes, such as instance family and size, location, OS, number of count, and also the specific term, then the billing will be automatically deducted by using the reserve instance. Reserve instance also supporting split, merge.

Here, we'll take a look on how does reserve instance works in this case. The customer or any purchaser regional reserve instance in Shanghai regions They purchase two units of general type of reserve instance and one unit of compute reserve instance. So during time to time, as long as we've been there, reserve instance term, that can be applied to offset the billing of the pay-as-you-go instance, which can match with specific attributes.

There are some limitation of the reserve instance, such as maximum number of reserve instance, each account can have up to 20 regional reserve instance in all regions and also 20 zonal reserve instance in each zone. So customer can raise ticket to purchase more reserve instance if they already reached the limit. Reserve instance cannot be used to offset the charges for network and storage resources of pay-as-you-go instance. The instance family of gn6i and t5 do not support for regional reserved instances and they only support zonal reserved instances. And both of these instances family are also cannot support the split and merged features.

In the next slide, we will take a look further on the saving plans detail. Saving plan is a cost optimization solution, which designed for long-term use of pay-as-you-go instances. Is is a new bidding method provided by ECS that allows users to enjoy pay-as-you-go billing discount of up to 77% off in exchange for an hourly commitment over one or three year period. So there are three major benefits that are provided by saving plans.

The first one is flexibility of purchase. Savings plan is available to purchase from one cent and up to no limit. Payment can be made in upfront or installments. The second benefit cost optimizations. Yearly discount at price level with a pay-as-you-go billing flexibility. And the third benefit is easy management. Saving plans will apply automatically and free of maintenance and applicable for both ECS and ECI instances across the regions, types and also accounts.

Next we'll take a look on the example on how does saving plan work? In this case, the customer will a 25 instance A which are running in the first hour and 15 instance are running in the second hour. So before we start the calculations, we will take a look on, customer already bought hourly commitment of USD $100. And the price rate, for instance A with saving plan is a $5 per hour. And the price rate for instance A for a normal pay-as-you-go-price rate, it will be $12 per hour.

So we will be focusing on first the pay-as-you-go billing, if the customer is purely using the pay-as-you-go method. So for the first hour, the calculation will be USD $12 times 25 unit of instance. So the total will be 300 USD for the first hour. For second hour, since already reduced to 15 instance, so the calculation will be 15 times $12 and the total will be 180 in the second hour. So the total of these two hours will be $480 USD. So now we will take a look on how saving plans help to reduce the cost.

So-called hourly commitment, USD $100, it can be covered on the 20 instance A, which the calculation will be a $100 divided by five, which is the saving plans price rate. And the remaining instance A, will be paying as a normal pay-as-you-go rate. So the total bill for the first hour will be $100 plus five times $12. So total will be 160 USD for the first hour. And for the second hour, the calculation will be more simple because the total instance, that's running in a second hour is less than 20 units. So it can be fully covered by using the saving plan without popping out any additional amount. So the second hour will be a $100 USD. And the total two hours amount is about 260 USD. That's a 45% saving compared to the normal pay-as-you-go pricing rate.

Now we will take a look on the saving plan payment options. So there are three main types of payment options. All upfront, 50 unfront and also the zero upfront. So for zero upfront, they will have a lesser discount and the charges, I mean, the payment will be made by monthly installments. For all upfront, the full payment is required during the purchase. However, you have the highest discount rate compared to the rest. There are two types of saving plan in this case.

The first one, general purpose. So for general purpose saving plan, it comes with a better flexibility. A single savings plan can be applied to all ECS and ECI instances globally. The required configuration, which including commitment amount, payment option, and also the durations. So the benefit included regional flexibility, instance family flexibility, instance type flexibility, operating system flexibility, and also the service flexibility. Whereas for the second, second type ECS compute. It will provide a better discount rate, but a single savings plan can be applied to the ECS and ECI instances of the specific instance families within a specific region. So the required configuration, including regions, purchase method, instance family, commitment hour, payment option, and also the durations.

So the benefit for ECS computes saving plan basically are global- I mean, better discount and also can be zone flexibility, instance type flexibility and operating system flexibility. Here is the saving plan implementations overview. So customer can see, I mean, can have the consumption predictions during the budgeting phase. So he will be able to predict an approximate hourly commitment based on the usage over the last seven, 30, or 60 days. And during the application phase, user will be able to allow using the saving plans across a month, among multiple accounts based on trusteeship. And the newer features are still under planning, which is the guaranteed resource supply.

The resource will be supplied and offset a fee for resource reservations. And this feature is still under planning. And doing the reconciliation phase, there will be a usage report available coverage report, and also the deduction details for customer to review and analyze their current spending. There are some limitations of the saving plan. Maximum up to 40 saving plans for each account. And saving plans cannot be applied to the instances of retire generations, such as t1, s1, s2 and et cetera.

ECS compute saving plans can be applied only to pay-as-you-go instance of a specific compute optimized instance family within a specific region. And saving plans are applicable to the following resources, such as vCPUs and memory. System disks and public bandwidth. The last but not least, we will have a quick overview on the cost optimizations method. In this slide, we will tell you, I mean, we will base on the different types of workload patters and there are some suggest- suggested purchase options for you to refer.

So for the first diagram, you see that it's kind of applied concurrency. So we will recommend customer to have saving plans plus the pass-as-you-go instance or the reserve instance with pay-as-you-go instance. For the second workload pattern is, is more like consistent. So customer can consider to use the long-term subscription plus pay-as-you-go or they also can consider for saving plan or reserve instance plus pay-as-you-go instance. And the number three, the workload pattern is more like a spiky type workload.

So in this case, the customer can consider to use the pay-as-you-go instance, which is more applicable. And they can also use the saving plan if the total discount overcomes the cost when resources are not needed. Lastly, multi workloads and mixed usage, customer can consider to use the saving plans plus pay-as-you-go instance, and also the reserved instance plus pay-as-you-go instance, if the resource are under the same instance family.

We also provide a comprehensive cost optimization solution, which including with other product discounts such as storage capacity units, SCU and data transfer plan. Storage capacity unit can be used to offset the pay-as-you-go bills of a variety of cloud storage services which including cloud disk, snapshots, OSS and NAS to streamline your service purchase process and the customer can buy from one TB up to 50 TB plan and is able to cover different regions and availability zones.

For data transfer plan, it is a discount plan for bulk usage discount and customer can purchase a data transfer plan that will be used to deduct on-demand per GB traffic usage, with this, all the other products, discount plans. Therefore customer can have a better cost optimization by leverage on this available options. That's all for my presentation today. Thank you for your time and hope this is fruitful to you. Thanks again.

About the Author
Learning Paths

Alibaba Cloud, founded in 2009, is a global leader in cloud computing and artificial intelligence, providing services to thousands of enterprises, developers, and governments organizations in more than 200 countries and regions. Committed to the success of its customers, Alibaba Cloud provides reliable and secure cloud computing and data processing capabilities as a part of its online solutions.