Cloud server ECS selection guide and money-saving magic weapon(strongly recommended collection)

Introduction:
How to choose a cloud server ECS and how to save money? This one is enough

The sharing brought to you today is how to buy cloud server ECS and how to buy more money, divided into four parts:
The first part introduces the basic concepts of ECS and tells you what parameters to look for when purchasing an ECS instance. Just like a young scholar who wants to buy a mobile phone, he will pay attention to the memory size, CPU frequency, screen resolution, camera parameters, etc. When purchasing an ECS instance, you can also choose parameters to meet your cloud requirements.

Main ECS product series of Alibaba Cloud in detail. (If you are short on time, you can skip to Part 3)

The third part explains in detail the selection skills of ECS, how to select ECS instances in different scenarios (such as big data/database, etc.), or what production and work scenarios an ECS instance is suitable for.

The fourth part introduces how to save money and effort to use ECS. On the premise of meeting your own needs, it will make it more economical to use the cloud to save money.

01Basic concepts of cloud server ECS


Basic Concepts of Cloud Servers
The first part will introduce some basic concepts of cloud servers to you.

Before we start, you can recall, what factors do we consider when buying a laptop ourselves? I will choose the brand first. Generally, after the brand is determined, the hardware configuration will be considered next, mainly the physical hardware configuration and software configuration.

In terms of hardware configuration, I will first consider the computing performance, such as the size of the CPU and memory, the model of the CPU, etc.; the second is the storage, how big is the disk of the laptop; For game students, the graphics card configuration is also very important. In addition to hardware configuration, I will also consider what the computer's operating system is, such as Mac OS, windows ubuntu, etc. After getting the computer, we will first install and configure some basic application software, including firewalls to ensure the security of our entire application environment.

This is the process of buying a physical computer in real life. In fact, these concepts are also applicable to the cloud. For example, when we choose some physical hardware parameters, we choose CPU and memory, which corresponds to the cloud. It is to select the CPU and memory size of the ECS instance and the model of the CPU.

For storage, the concept corresponding to disk on the cloud is block storage. Block storage on the cloud actually includes two concepts, one concept is a cloud disk, and the other is a local disk. There is a point that is different from our real life, which is the block storage on the cloud. During the purchase process, we need to specify whether it is used as a system disk or a data disk. In real life, if you buy a computer, there is a disk in it, and then we will divide the disk into a system disk or a data disk, but the system disk and data disk on the cloud need to be purchased separately, which is a little bit different.

The network is actually similar. An elastic network card is provided on the cloud, allowing users to connect to the Internet by accessing the cloud server.

In addition to this physical hardware, to make a cloud server really run, just like in real life, we also need to install an operating system. The concept of this operating system on the cloud is an image. Alibaba Cloud provides a variety of different image versions. For everyone to choose.

In addition, cloud servers also have some special concepts, such as security groups, which essentially limit the access traffic through some rules, that is, which applications can access.

After we buy a computer, this physical machine is in hand, and you can use it whenever you want. After buying a cloud server on the cloud, because the server is in the cloud or remotely, the way we access the cloud server is not the same as when we usually open a computer. We need to go through the Alibaba Cloud console or remotely. connected tool to log in to our cloud server.

Another small concept is the disaster recovery backup capability on the cloud, that is, snapshots. In real life, if our computer disk fails and the data is damaged, there is nothing we can do, or we can only find a professional person to get the data back, but there is no guarantee that all the data can be recovered. There is a concept of snapshots on the cloud. It means that taking a picture of the data at a certain point in the cloud disk is essentially recording all the data on the disk. If there is a problem, we can pass Snapshots can be quickly rolled back to data at a certain point in time, which can ensure rapid disaster recovery in the event of business problems .

In the overall introduction of the basic concepts of cloud servers, let's introduce the concepts of cloud server storage and network in detail.
Three storage methods on the cloud

The first is the block storage mode that has been introduced earlier. After the user creates block storage, he can mount the block storage to the instance, just like the process of using the laptop himself, if the disk that comes with the computer is not enough, go to Buying a portable hard drive to plug it in is similar. There are three types of block storage, including ordinary high-efficiency cloud disks, SSD cloud disks, and ultra-high-performance ultra-low-latency ESSD cloud disks.

The second storage method is file storage. Each block storage can only be mounted on one cloud server, and each file storage can be used by multiple ECS.

The third storage form is the object storage form OSS, which is similar to the Baidu cloud disk. Using this storage method, more files are read through a link.

Network on cloud

The network part is mainly two concepts, VPC, and switch.

The first is the VPC, which is a private network that divides a private network for users on the cloud. By creating a VPC, users can create a network environment that is completely logically isolated. Each VPC consists of a router and more than one composed of switches. Once a user creates a VPC private network, Alibaba Cloud will automatically create a corresponding router for the user to complete the forwarding of all networks under the VPC. The intranets between instances in the same VPC are interconnected, that is, instances in the same VPC can access each other through intranet IP addresses.

The second concept is the switch, which has been introduced earlier. A VPC has at least one router. The switch is the basic network device of the VPC, which is used to connect different instance resources. We can create multiple switches in each availability zone through the switch to divide the subnet, and then the multiple switches can be connected through routers. and forwarding. The above are some basic concepts of storage and networking.

The use process of cloud server ECS:

Below we introduce the process of using ECS.

An ECS instance, we can understand as a virtual machine, which includes software and hardware such as memory, disk, network, and operating system. The size of an ECS server instance and the underlying physical hardware are determined by the corresponding instance type and instance type family. The instance type family represents the applicable business scenarios of the instance, it determines the CPU and memory ratio, and what the underlying physical hardware looks like. The instance specification represents the size of the instance, such as the number of CPUs.

After determining the instance specifications, we also need to select the corresponding storage, because if there is only CPU and memory, there is no way to store data, so there will be block storage. There are two types of block storage, one is a cloud disk and the other is a local disk. A cloud disk is actually a three-copy storage form on the cloud, which can provide users with high-availability capabilities. Cloud disks are mainly used as system disks and data disks. You only need to format them like physical disks before use, while local disks may be mainly used as data disks.

After choosing the computing storage, we need to look at the corresponding operating system. The operating system on the cloud refers to the image. At present, Alibaba Cloud provides a variety of image sources, including the official public image and the third-party market. Mirrors are user-defined mirrors, and also allow sharing of mirrors between different users.

In terms of network, Alibaba Cloud will have a network bandwidth that users can directly specify.

After we define the parameters of the instance's computing, storage, network, and operating system, we can create a cloud server that is the same as our physical laptop.

After the creation, we can directly connect and access the purchased cloud server through the Alibaba Cloud console or through the Alibaba Cloud APP.

Introduction to Cloud server ECS instance type family

In the second part, I will introduce to you how the specification families of ECS instances are named. You may have many questions in this section. At present, Alibaba Cloud provides hundreds of instance specifications, so you will be dazzled in the selection process. In fact, as long as you understand the naming method of the ECS instance specification family and its information layout, we can make a good selection.

Instance's architecture type, specification classification, and details

On the purchase page of the Alibaba Cloud console, you can see that the selection of instance type families is divided into three modules: architecture, classification, and specific information. The top is the type of our instance specification architecture. There are three types of architectures, namely the general X86 architecture, heterogeneous computing (such as GPU or FPGA, NPU, etc.), and Alibaba Cloud's self-developed Shenlong bare-metal architecture.

There will be a classification of instance specifications under each architecture. As can be seen from the above figure, under this computing type of X86, there are 7 types of instance specifications.

Different instance specifications represent different hardware configurations. Choose any instance. After the specifications are classified, we can see the detailed information of the corresponding instance specifications, which are mainly divided into four parts:

The first is detailed information about the instance type family, including the corresponding type family and the name of the instance type. Here, you can click the small question mark to see some detailed descriptions of the instance type family.

The second part is the information on CPU and memory size, here is what everyone will pay more attention to in the selection process.

The third part is the network capability information of the instance, including the bandwidth of the instance's intranet and the ability to send and receive packets.

The fourth part is the information on the processing model of the CPU, including the main frequency and the turbo frequency of the processor.

Enterprise Instance vs Entry Instance

As you can see on the purchase page of the console, there are many instance type families of ECS, and it is impossible to judge the difference between them simply from the CPU and memory, so we need to look at it from a macro perspective. The instance specifications of Alibaba Cloud ECS are divided into two categories: one is an enterprise-level instance, and the other is an entry-level instance.

The enterprise-level instance was launched by Alibaba Cloud in September 2016. Its feature is that vCPU is exclusive, which means that when we create an enterprise-level instance, the instance vCPU is bound to our underlying physical CPU. , the underlying physical CPU can no longer be allocated to other instances, so enterprise-level instances will not compete for resources, so stable performance can be guaranteed, and enterprise-level instances provide very strict SLA performance guarantees.

The entry-level instance is that the CPU is not bound to the underlying physical CPU, which means that each vCPU may be randomly assigned to an idle physical CPU at the bottom. If there are multiple shared entry-level instances on the same physical server If the instance is at the same level, different instances will compete for resources, resulting in unstable CPU performance.

Because entry-level instances have unstable performance, Alibaba Cloud only provides one entry-level instance, which is a shared instance in the X86 architecture, and other instance specifications in the X86 architecture, as well as in the heterogeneous architecture and the Dragon architecture. All instances are enterprise-class instances.

Because enterprise-level instances have stable performance and strict SLA guarantees, they are more suitable for scenarios that require high business stability. Since entry-level instances cannot guarantee performance stability, they are relatively cheap and are more suitable for scenarios that do not have strict performance requirements or have performance burst requirements during certain periods of time, such as some light-load applications or Microservices.

shared instance

After introducing the large classification of ECS instances, let's take a look at the specific information of shared instances.

We mentioned earlier that only shared instances under the X86 architecture are entry-level instances. This type of instance has one more parameter than the previous instance, which is "average benchmark CPU computing performance", and the benchmark performance is the CPU performance that the instance can continuously provide.

Shared instances, also known as entry-level instances, are divided into two categories. The first category is a standard shared instance. The CPU is not bound and only provides the benchmark CPU performance. Therefore, when there is competition for resources, can it exceed the benchmark? Performance is not guaranteed.

Another special shared instance, named burst performance shared instance, is mainly to take care of the fact that some applications may not have high CPU usage and low load most of the time, but in certain applications Sometimes there may be temporary burst performance requirements, so Alibaba Cloud will provide burst performance parameters, so when you purchase a shared instance, you can obtain higher than average benchmark CPU performance through burst performance. ability.

For a shared instance with burst performance, if the actual usage of the application is lower than the average benchmark performance, the corresponding CPU credits will be obtained. If the performance requirement suddenly increases in some scenarios, for example, the usage rate of the corresponding CPU of the instance exceeds 20% will consume the previously accumulated CPU points to improve computing performance so that computing performance will not be affected. This is a unique feature of shared instances with burst performance.

Two special instance sizes

In addition to shared entry-level instances, Alibaba Cloud also has two instances with special specifications, namely big data and local SSD.

These two instance specifications will come with local storage. The local storage of big data instances is an HDD disk, and the new local storage of local SSD is a local SSD disk with very high I/O throughput and low latency. The specific information can view it in the Alibaba Cloud console.

Enterprise Instance Spec Family Tree

The following introduces the family tree of enterprise-level instance specifications so that we can quickly understand the "kinship" relationship of each instance family. The enterprise-level instance family is divided into three parts. The first block is X86 computing. In addition to the shared type, the general-purpose, computing, memory, high-frequency, local SSD, and big data types all belong to our enterprise-level instances. Instances are constantly iterated every year, so they will be divided into different generations, and we will detail the differences between different generations later. All GPUs and FPGAs in heterogeneous computing are enterprise-level instances, as are bare metal and high-performance computing.

First, let's introduce the naming methods of X86 instance specifications, which are divided into 5 types:

The first instance type is the general-purpose type. As the name suggests, it can be used in any scenario. Therefore, this type of instance is called the g-series. Its CPU and memory ratio is 1:4.

The second instance type is the computing type. As the name implies, the CPU computing power requirement will be higher in some scenarios. Therefore, its vCPU and memory ratio is 1:2 and is then referred to as the c series.

The third type is the memory type, which provides more memory capacity, so its CPU and memory ratio is 1:8, also referred to as the r series, and r is the abbreviation of RAM.

The fourth and fifth types are big data type and local SSD type respectively. The ratio of CPU and memory of these two types is 1:4, but the types of local disks they are equipped with are different, which leads to their skills. It is also different from the suitable scene. So the abbreviation for big data type is d, and the abbreviation for local SSD type is I .

On these five basic instance specifications, we will do some additional capacity enhancements. For example, under the three types of general-purpose, computing, and memory, we have added some high-frequency capabilities. The main frequency should be 2.5G Hz, but some of us can achieve 3.2G Hz. This ability to add a high main frequency becomes a high main frequency type, and a logo like hf will be added in front of it.

With the evolution of technology, the Shenlong card with the Shenlong architecture is also constantly iterating and improving. The Shenlong card equipped with the third generation can improve the performance of the three instance specifications of general-purpose, computing, and memory as a whole, so there will be A balance enhancement. For the big data type, the separation of computing and storage is done, forming a big data storage type, referred to as d2, and d2s is based on big data, making some network capabilities enhancement, it becomes a network Enhanced.

Naming method and rules of instance specifications

You can see the naming method and rules of Alibaba Cloud instance specifications in the following figure.

The normal X86 instance specification name is divided into three sections. The first part indicates the product name, and ECS is a product of Alibaba Cloud. The second part indicates the instance specification and generation. As mentioned earlier, hfg indicates that it is a general-purpose type. On the basis, the ability of high frequency is added, and then what does 6 represent? In fact, it represents the generation of our products. You can calculate the new or old of the corresponding product according to the generation of the product. For example, 6 represents the 6th generation, and 5 represents the 5th generation. The larger the number, the newer it is. For a generation, the underlying physical hardware will be newer, and its price/performance ratio will be relatively higher.

The last part is the specification of the instance, which indicates the number of vCPU cores of the instance. Large represents 2 vCPUs, xlarge represents 4 vCPUs, 2xlarge represents 8 vCPUs, and so on.

Knowing the above naming rules, you can use the name of the instance type family to infer what type of CPU the current instance has, what generation it has, and the number of CPUs it has.

The GPU naming rules are also similar, with only one difference. The middle part of the GPU name will provide the ratio between the CPU and the GPU because the GPU is a card that provides an additional GPU in addition to the CPU. Therefore, we can also directly infer its underlying physical configuration through the format of its specification family.

03Cloud server ECS instance selection in practice

The third part will tell you how to do the selection of cloud server ECS.

Briefly describe the applicable scenarios of various specification instances

X86 calculation:
X86 general-purpose, computing and storage instances have the same ratio of CPU and memory, so they are more suitable for small and medium-sized databases or some data processing tasks.
The c series is mainly computational, so it is more suitable for some computing requirements, such as some external applications, batch computing, or high-performance scientific computing. • For the r series, because it has more memory, it is more suitable for some database or data analysis applications. High-frequency instance specifications are also more suitable for high-performance scientific computing that requires relatively high CPU frequency. • The local SSD type is more suitable for some relational databases or NoSQL databases. • The D-series big data type may be more suitable for some big data cluster scenarios, such as this Map Reduce kind.

In this part of heterogeneity, it is divided into two categories:
GPU is more suitable for deep learning or image and video visualization processing;
FPGA is more suitable for image transcoding, or audio and video decoding.

Bare Metal and HPC:
Some scenarios with more vertical and higher performance requirements, such as some high-performance databases or high-performance scientific computing scenarios.

Below we give a few examples to introduce the selection method in detail.

X86 instance selection recommendation

We can divide a web application into the following levels, and make corresponding recommendations for each level:
• For Apache and Nginx web servers, because it mainly does some computing processing, it is recommended to use some computing types, such as c5, c6;
• For middleware such as spring cloud or MQ, it belongs to the normal requirements for computing and storage, so we recommend some general-purpose, such as g6 instance specifications; • The application type is a relatively general scenario, so the G6 series can meet it; • Cache applications such as Redis and Memcache have relatively high memory requirements, so we recommend using in-memory types, such as r series;
For relational databases, we can directly use in-memory types, such as r series with our • For NoSQL, we recommend local SSD type, such as I series; For big data, similar to HDFS or spark, we also have a special big data type, such as D series To deal with; • For the bottom-level machine learning, such as MXNet training framework, there will be a corresponding special GPU computing type.

Recommended GPU instance selection

The scenarios of GPU cloud servers are mainly divided into two categories. The first category is artificial intelligence, or machine learning, and the second category is graphics and image processing. In machine learning, it will also be divided into two scenarios, one is training and the other is inference. Therefore, for the vertical fields of different subdivisions, we have given some recommended specifications, which can be seen in the following figure.

Below we introduce two relatively complex selection scenarios.

Example selection of big data scenarios

The first complex scenario is the big data scenario. When building a big data cluster similar to Hadoop and Spark, if we build it manually, the process will be divided into three major parts: the first major part in the management node of the cluster. In instance specification selection, the second block is the selection of cluster computing nodes, and the third block is the selection of cluster data nodes.

The management node is a common scenario, so directly selecting the g series can handle the management tasks well;

The computing node is more of a normal business load, so the g series can be used as the main choice. SSD cloud disk; • Data nodes have relatively high requirements on storage throughput and network throughput, so it is recommended to use the d series, with the corresponding local disk, to complete the read volume of this data;

Therefore, on the whole, in the same big data cluster, different tasks have different characteristics, so different instance sizes will be selected.

Database scenario instance selection practice

The second complex scenario is about database selection:
For ordinary business, databases with a relatively light load can be handled by special general-purpose g-series or in-memory r-series with high-efficiency cloud disks and SSD cloud disks, resulting in higher cost performance. Because the price of the g series is still very advantageous compared to local disks or local SSDs. The overall performance of high-efficiency cloud disks and SSD cloud disks can actually meet the daily database scenarios.
For clusters with very high business load requirements, the i -series cloud disks equipped with NVMe SSDs with local SSDs are recommended, which can achieve high storage IOPS and low latency and can meet the performance requirements of heavily loaded databases.

X86 Gen 6 vs Gen 5 instance price comparison

In addition to performance, everyone will also pay attention to the price. Here is a price comparison between the 6th generation and the 5th generation in the X86.

It can be seen that in some regions, the price of the 6th generation instance is slightly higher than that of the 5th generation 10 instance, and the monthly price of general-purpose and memory types is generally higher than that of the 6th generation. The 5th generation is 2%-12% cheaper, so overall, the 6th generation is not only a 20% improvement in performance but also the vast majority of products will be cheaper.

If you pay by volume, the price of the 6th generation will be 37%-47% lower than the price of the 5th generation, which is actually a very large room for profit. So when choosing to buy ECS by volume, choosing the 6th generation will be cheaper than the 5th generation.

Selection and combat summary

To summarize the selection method, there are three rules, which you can keep in mind and use in the selection process.

The first rule is that enterprise-level instances of the same size have more stable performance than entry-level instances, but entry-level instances are more cost-effective because enterprise-level instances have exclusive vCPUs, and there is no competition for resources. Performance guarantee, but for some personal or small and medium-sized website applications, if the performance requirements are not so strong, choosing an entry-level instance is actually a better choice.

The second rule is that under the same instance specifications, the new generation of instance specifications is more cost-effective than the old generation. Users release more technical dividends.

The third rule is not only to choose the appropriate instance size but also to carry the appropriate block storage so that the application on the cloud can achieve the expected performance. The cloud will provide four different types of block storage, including high-efficiency cloud disks, SSD cloud disks, ESSD cloud disks, and local disks. The IOPS and throughput of different types of disks are different. It is also necessary to choose the appropriate block storage in order to form a joint force and achieve the best performance.
The third part will tell you how to do the selection of cloud server ECS.

Briefly describe the applicable scenarios of various specification instances

X86 calculation:
X86 general-purpose, computing and storage instances have the same ratio of CPU and memory, so they are more suitable for small and medium-sized databases or some data processing tasks.
The c series is mainly computational, so it is more suitable for some computing requirements, such as some external applications, batch computing, or high-performance scientific computing. • For the r series, because it has more memory, it is more suitable for some database or data analysis applications. High-frequency instance specifications are also more suitable for high-performance scientific computing that requires relatively high CPU frequency. • The local SSD type is more suitable for some relational databases or NoSQL databases. • The D-series big data type may be more suitable for some big data cluster scenarios, such as this Map Reduce kind.

In this part of heterogeneity, it is divided into two categories:
GPU is more suitable for deep learning or image and video visualization processing;
FPGA is more suitable for image transcoding, or audio and video decoding.

Bare Metal and HPC:
Some scenarios with more vertical and higher performance requirements, such as some high-performance databases or high-performance scientific computing scenarios.

Below we give a few examples to introduce the selection method in detail.

X86 instance selection recommendation

We can divide a web application into the following levels, and make corresponding recommendations for each level:
• For Apache and Nginx web servers, because it mainly does some computing processing, it is recommended to use some computing types, such as c5, c6;
• For middleware such as spring cloud or MQ, it belongs to the normal requirements for computing and storage, so we recommend some general-purpose, such as g6 instance specifications; • The application type is a relatively general scenario, so the G6 series can meet it; • Cache applications such as Redis and Memcache have relatively high memory requirements, so we recommend using in-memory types, such as r series;
For relational databases, we can directly use in-memory types, such as r series with our • For NoSQL, we recommend local SSD type, such as I series; For big data, similar to HDFS or spark, we also have a special big data type, such as D series To deal with; • For the bottom-level machine learning, such as MXNet training framework, there will be a corresponding special GPU computing type.

Recommended GPU instance selection

The scenarios of GPU cloud servers are mainly divided into two categories. The first category is artificial intelligence, or machine learning, and the second category is graphics and image processing. In machine learning, it will also be divided into two scenarios, one is training and the other is inference. Therefore, for the vertical fields of different subdivisions, we have given some recommended specifications, which can be seen in the following figure.

Below we introduce two relatively complex selection scenarios.

Example selection of big data scenarios

The first complex scenario is the big data scenario. When building a big data cluster similar to Hadoop and Spark, if we build it manually, the process will be divided into three major parts: the first major part in the management node of the cluster. In instance specification selection, the second block is the selection of cluster computing nodes, and the third block is the selection of cluster data nodes.

The management node is a common scenario, so directly selecting the g series can handle the management tasks well;

The computing node is more of a normal business load, so the g series can be used as the main choice. SSD cloud disk; • Data nodes have relatively high requirements on storage throughput and network throughput, so it is recommended to use the d series, with the corresponding local disk, to complete the read volume of this data;

Therefore, on the whole, in the same big data cluster, different tasks have different characteristics, so different instance sizes will be selected.

Database scenario instance selection practice

The second complex scenario is about database selection:
For ordinary business, databases with a relatively light load can be handled by special general-purpose g-series or in-memory r-series with high-efficiency cloud disks and SSD cloud disks, resulting in higher cost performance. Because the price of the g series is still very advantageous compared to local disks or local SSDs. The overall performance of high-efficiency cloud disks and SSD cloud disks can actually meet the daily database scenarios.
For clusters with very high business load requirements, the i -series cloud disks equipped with NVMe SSDs with local SSDs are recommended, which can achieve high storage IOPS and low latency and can meet the performance requirements of heavily loaded databases.

X86 Gen 6 vs Gen 5 instance price comparison

In addition to performance, everyone will also pay attention to the price. Here is a price comparison between the 6th generation and the 5th generation in the X86.

It can be seen that in some regions, the price of the 6th generation instance is slightly higher than that of the 5th generation 10 instance, and the monthly price of general-purpose and memory types is generally higher than that of the 6th generation. The 5th generation is 2%-12% cheaper, so overall, the 6th generation is not only a 20% improvement in performance but also the vast majority of products will be cheaper.

If you pay by volume, the price of the 6th generation will be 37%-47% lower than the price of the 5th generation, which is actually a very large room for profit. So when choosing to buy ECS by volume, choosing the 6th generation will be cheaper than the 5th generation.

Selection and combat summary

To summarize the selection method, there are three rules, which you can keep in mind and use in the selection process.

The first rule is that enterprise-level instances of the same size have more stable performance than entry-level instances, but entry-level instances are more cost-effective because enterprise-level instances have exclusive vCPUs, and there is no competition for resources. Performance guarantee, but for some personal or small and medium-sized website applications, if the performance requirements are not so strong, choosing an entry-level instance is actually a better choice.

The second rule is that under the same instance specifications, the new generation of instance specifications is more cost-effective than the old generation. Users release more technical dividends.

The third rule is not only to choose the appropriate instance size but also to carry the appropriate block storage so that the application on the cloud can achieve the expected performance. The cloud will provide four different types of block storage, including high-efficiency cloud disks, SSD cloud disks, ESSD cloud disks, and local disks. The IOPS and throughput of different types of disks are different. It is also necessary to choose the appropriate block storage in order to form a joint force and achieve the best performance.

04 The way to save money and effort with Cloud server ECS

When purchasing a cloud server, in addition to selecting the instance specification to make the selected instance specification more closely matched with the business, we also need to consider whether it can be cheaper and whether the resource delivery can be completed quickly. The last part introduces you to the skills of ECS to save money and effort.
save money

The first is to save money, which means that after selecting the instance specification, you need to choose the most suitable payment method to get better cost optimization. Alibaba Cloud currently provides 7 payment methods, such as Saving Plan, Annual Subscription, Reserved Instance Voucher, Pay-As-You-Go, and Preemptible Instance.

How to choose the right payment method? There is a strategy, that is, we need to choose the most suitable payment method according to the stability of the business and the fluctuation of peaks and valleys.

For example, saving plans, annual subscriptions, and reserved instance coupons are more suitable for stable business loads; for stateful and dynamically changing business loads, you can use pay-as-you-go; for completely stateless and high capacity Disaster-resistant, preemptible instances can be used for delivery, because the price of preemptible instances can be 10% of that of pay-as-you-go instances.

The way to save effort

The second is the way to save effort. When purchasing resources on the cloud, sometimes bulk purchases are made. Alibaba Cloud provides a variety of automated resource delivery modes and tools, enabling one-time configuration and reuse, thereby improving the speed and efficiency of deployment on the entire cloud .

For example, batch delivery is done through the console; the delivery of the underlying computing power cluster with disaster tolerance capability can be completed through the deployment set; the delivery of resources can be completed automatically through elastic scaling and elastic supply; and through resource orchestration, various Different combinations of resources are delivered.

Four steps for cloud model selection

To sum up, in the process of migrating to the cloud, we need to take four steps:
Step 1: Do some analysis on your own business characteristics, including performance requirements and network requirements, to form a basic judgment;
Step 2: Select the corresponding ECS instance specification according to the service characteristics; Step 3: Select For a corresponding payment method, only by choosing the most appropriate payment method can we achieve the optimal cost on the cloud ;
Step 4: Choose an appropriate delivery method to help us complete the delivery of resources in a time-saving and labor-saving manner.

Related Articles

Explore More Special Offers

  1. Short Message Service(SMS) & Mail Service

    50,000 email package starts as low as USD 1.99, 120 short messages start at only USD 1.00