From AWS to Alibaba Cloud: Analysis of Product System Differences

Introduction: With the increasingly fierce competition in the international cloud market and the Sino-US trade war as a fuse, more and more domestic AWS customers have begun to migrate from AWS to Alibaba Cloud. This article, as the beginning of a series of articles, analyzes the differences between AWS/Alibaba Cloud products as a whole, hoping to give you some lessons and inspiration aws vs alibaba cloud pricing.

calculate
Although some products are different in function, the overall product system is not much different between AWS and Alibaba Cloud .

AWS has made big mistakes in product direction. In terms of container products, AWS has concentrated on Elastic Container Service before, and launched the kubernetes product Fargate in 2018.

The benchmarks of common computing products of AWS/Alibaba Cloud are shown in the following table aws vs alibaba cloud pricing.


Table 1: Common Computing Products of AWS/Alibaba Cloud
Taking EC2/ECS as an example, the following table compares ECS/EC2 instances by type. The richness of instance types is basically the same. The main advantage of AWS is network performance. The maximum performance of a single instance can reach 100G. And there are technical advantages.


Table 2: Comparison of AWS/Alibaba Cloud Instances by Type
In addition, AWS's RI (Reserved Instance) has no restrictions on instance specifications, that is, calculated based on the total number of CPU and memory, Alibaba Cloud Reserved Instances have restrictions on specifications and specifications, and the flexibility is slightly less than AWS.



network
The following table shows the benchmarking of commonly used network products aws vs alibaba cloud pricing.

It is worth mentioning that the CEN product was first released by Alibaba Cloud, and the AWS TGW product is about a year later than Alibaba Cloud CEN.


Table 3: AWS/Alibaba Cloud Common Network Products
From a functional point of view, AWS's load balancing products support more forwarding strategies than Alibaba Cloud SLB, as shown in the following table.


Table 4: Function comparison of commonly used network products in AWS/Alibaba Cloud


storage
The following table shows the benchmarking of commonly used storage products.


Table 5: AWS/Alibaba Cloud Common Storage Products
Taking S3/OSS as an example, analyze the differences between S3/OSS from the following points aws vs alibaba cloud pricing

API Compatibility:

Access OSS via S3 API

Obtain the AccessKeyId and AccessKeySecret of the OSS main account or sub-account, and configure the requested AccessKeyId and AccessKeySecret in the client and SDK
Considering OSS feature support and operation and maintenance, it is recommended to use the OSS SDK aws vs alibaba cloud pricing

Path style and Virtual hosted style: virtual hosted style refers to the access method that puts the Bucket in the host header. Based on security considerations, OSS only supports virtual hosted access. Therefore, after the migration from S3 to OSS, the client application needs to be configured accordingly. Some S3 tools use Path style by default, and you need to configure them accordingly. Otherwise, OSS may report an error and prohibit access. Subsequent S3 will also gradually remove the Path style method
Differences in public VPC access: Access to S3 on the public network or within the VPC uses the same Endpoint, and access within the VPC is equivalent to private use of the public network address. OSS needs to use different endpoints. Considering the migration speed and cost, S3 migration to OSS generally adopts the private line method, and the migration service needs to be built in the AWS EC2 cluster;
Multi-version: Only some regions of OSS have multi-version enabled
API compatibility: S3/OSS compatible APIs, as shown in the following table:

Table 6: S3/OSS Compatible APIs


database
The following table shows the benchmarking of commonly used database products.


Table 7: AWS/Alibaba Cloud Common Database Products
On the whole, the database product systems are relatively similar, and the major differences are:

AWS does not have a native MongoDB product, but has launched DocumentDB, which is compatible with MongoDB API and has a separate architecture for storage and computing;
AWS does not have a native HBASE product, you can quickly use HBASE through EMR; Alibaba Cloud's EMR also supports HBASE;
AWS supports native Oracle, Alibaba Cloud is partially compatible with oracle through POLARDB-O, PPAS, and goes to O evaluation in combination with ADAM products
Take RDS as an example to introduce the advantages and disadvantages of AWS/Alibaba Cloud:

AWS/Alibaba Cloud commonalities

Managed database service with good scalability
Supports relational databases such as MySQL, PostgreSQL, SQL Server, etc.
High availability deployment (multi-availability zone), Failover automatic switching
data encryption, security
Reliable file storage (based on cloud disk/EBS)
AWS/Alibaba cloud difference

AWS RDS implements disaster recovery through dns, and IP changes are triggered during active/standby switchover and up/down configuration; Aliyun RDS introduces proxy links to achieve read/write separation, seven-layer load balancing, and avoid continuous flash interruptions. Does not trigger IP changes
Version 6 read-only instance does not support Binlog
Aliyun supports VPC & Classic; AWS only supports VPC
The default mysql system parameters are partially inconsistent
Alibaba Cloud Advantage

Better database performance, AWS RDS is deployed based on EC2, Alibaba Cloud is also promoting ALL ON ECS, and RDS has released the ECS version, the measured IOPS/QPS of Alibaba Cloud non-ECS version RDS under the same configuration is twice as high as AWS RDS , with an exclusive physical machine instance that AWS does not have
Supports asynchronous and semi-synchronous replication
Longer mirroring, log retention time
Cross-region synchronization based on DTS
SQL Audit
Combined with DMS, a better operation and maintenance experience
AWS Advantage

Native support for Oracle
More database parameters
AWS has a delete protection switch to avoid accidental deletion


Big Data
The following table shows the benchmarking of commonly used big data products.


Table 8: AWS/Alibaba Cloud Commonly Used Big Data Products
There is a big difference in the planning of AWS and Alibaba Cloud big data products, especially in terms of self-research, Alibaba Cloud's MaxCompute+DataWorks has no products that can match it in terms of functions.

Redshift is mainly aimed at the traditional data warehouse field, providing fast query and analysis of large-scale data. Redshift does not directly support MR. It can be combined with EMR for calculation and use Redshitf storage. As for IDEs, AWS does not provide native product support, instead recommending that its customers use open source IDEs.

The following describes the differences between Redshift and MaxCompute.


Table 9: Comparison of differences between Redshift and MaxCompute
In addition, AWS's EMR+S3 is much higher than Alibaba Cloud in terms of network performance. A single cluster can reach 500G bandwidth, while on Alibaba Cloud, EMR+OSS can only reach 100G natively. Combined with Jindofs, it can achieve similar 500G bandwidth. throughput. From this aspect, AWS's advantages in network infrastructure technology can also be reflected.



Summarize
With the rapid development of Alibaba Cloud in recent years, the product system has become more and more perfect, catching up with the footsteps of AWS, and some products have the first-mover advantage, such as CEN, and a few of our products are very advantageous For example, SLS, there is no product that can be compared with SLS on AWS. In terms of logs, AWS construction customers use the open source ELK solution (ElasticSearch can use AWS products); in addition, the domestic live broadcast industry in Benefit Country is hot, and Alibaba Cloud CDN live broadcast is more mature than AWS .

But overall, AWS is still the leader, and its advantages are mainly reflected in the following points:

Product consistency: From the actual experience of AWS users, all AWS products are predictable, standardized and systematic, and there will be no surprises or surprises. Reception horizontal characteristics or horizontal product docking must be relatively complete. However, many products of Alibaba Cloud are imperfect in these aspects. For example, many of our new products or newer products have imperfect labeling functions. The monitoring of migration/synchronization tasks of DTS products is configured in the product's own console. Access to cloud monitoring, etc., there are many such situations;
API concept: API concept brought by AWS e-commerce, product release strictly controls API; AWS API and documentation are relatively complete, our API is not stable enough, especially involving some new functions or multi-product collaborative API, there are many bugs , the documentation is not perfect
Ecology: The AWS ecosystem is relatively strong. For example, terraform is supported by ecological developers and developed by Ali's internal personnel.
Based on the stable API of AWS, mainstream foreign AWS users and domestic TOP customers have a very deep understanding and use of cloud, which is mainly reflected in the following two points:

Ultimate elasticity:
AWS RI (Reserved Instance) does not limit ECS specifications. Reserved Instances can be used for different services at any time as needed, which can be online or offline applications;
Online applications can be stably scaled according to business peak and low peak times or business real-time traffic curves, keeping the CPU usage of online applications above 30%;
Offline platform EMR: The cluster can be released directly during the day, and the offline tasks can be created regularly at night
Infrastructure as Code:
Completely out of the console, based on API operations, using Terraform to manage cloud resources stably and efficiently
Creating a k8s cluster on the cloud (based on products) requires 20+ steps for console operations, and errors may occur. Based on Terraform, when the configuration is written, it can be created in one step, providing assistance for extreme elasticity


The future of cloud
Standardization: The cloud in the future is bound to become more and more standardized. The underlying implementation technologies may be different between different vendors, but the functions, protocols, and interfaces provided to users are more and more consistent.

Barriers: Under the trend of standardization, the barriers of different cloud vendors are more in basic technology, product quality, and user experience, providing users with better performance, more stable product services, and better user experience.