Price comparison of six cloud vendors - Alibaba Cloud

In August last year, we selected six cloud manufacturers at home and abroad to do a resource price report, and then released the "Cloud Toolkit 2.0" aws vs azure vs google cloud.

It's been almost a year and we have a new one aws vs azure vs google cloud.

The most important thing is to have a beginning and an end.

This year, we still selected the six public cloud vendors AWS, Alibaba Cloud, Azure, and Google Cloud for overall comparison. The comparison includes the following four items:

Overall situation: geographical distribution/billing model/resource type aws vs azure vs google cloud;

Four resource types: computing CPU/general-purpose CPU/memory CPU/GPU;

Two billing models: On-Demand Instances/Reserved Instances (monthly/1 year/3 years);

Four types of areas: optimal area/Beijing area/Shanghai area/US area aws vs azure vs google cloud.

Compared with last year, there are three main changes:

1. The overall situation and prices of major cloud manufacturers have been fully updated to the June 2020 version, and compared with the same period last year. We have selected similar resources for re-comparison of resources that have been offline;

2. Added memory CPU to the resource type aws vs azure vs google cloud;

3. Added Shanghai region.

As a whole, we still follow the price comparison principle of "comparing prices without comparing services, not considering preferential promotions, and trying to find similar performance resources". For specific settings, please refer to this article last year.

We will update the complete updated version of the price comparison report and the original Excel data document to "Cloud Toolkit 3.0". If you are interested, you can contact Xiao F at the end of the article to get it.

2020 Update Conclusion
Let's talk about the conclusion of the 2020 update:

First: major cloud vendors are expanding their territories and building new data centers around the world.

1. The increase in the China region in the past year is mainly reflected in Alibaba Cloud.

2. Four cloud vendors chose to build new data centers in Asia (excluding China).

3. The largest number of new regions in the world is Azure.

4. Domestic manufacturers still only have footprints in South America and Africa.

In April this year, Alibaba Cloud announced a three-year investment of 200 billion yuan in core technology and data center construction. In May, announced a five-year investment of 500 billion yuan, including cloud computing and large data centers and others.

Next year, we will wait and see.

Second: The instance types of major cloud vendors have not changed much, and only fine-tuned some names.

Third: The billing models of the six cloud vendors have barely changed. Only reserved instances are subdivided into months, 1 year, 2 years, 3 years, 4 years, 5 years.

Azure is operated by 21Vianet in China, and there is no direct open purchase channel for reserved instances, but similar discounts to reserved prices are achieved through the Azure CPP service.

Fourth: Based on the samples we took, the main price changes are concentrated on general-purpose CPU and GPU resources, and the areas with the greatest impact are mainly the optimal area and the United States.

The optimal area is defined here as the area where the global lowest price of a certain type of resource of a certain manufacturer is located in the recent period.

The overall changes in resource prices of various cloud vendors are as follows:

The price of general-purpose (standard S2) in the United States (Virginia) has been greatly reduced, and the 3-year reserved price has been reduced by more than 50%, but the price is still higher than that in mainland China.

The optimal areas for general-purpose products are in Chengdu and Chongqing, which are down about 10% compared to last year, and the prices of other products are basically unchanged.

Alibaba Cloud: Steady and rising

The price changes of Alibaba Cloud products were small, and the optimal regional price of its general-purpose product (ecs.g5.large) was slightly reduced by 1 point, located in Mumbai.

On-demand and reserved prices for US regional GPU products (gn5) have increased by 10%-15%. Other product prices remain unchanged.

Google Cloud: Slightly up

The on-demand prices of 2 of Google Cloud's 4 products in the US region (which is also the optimal region) (general-purpose n1-standard-2 and GPU NVIDIA Tesla P4) have all increased, but the price is not large. 3% or so.

Azure: Reservations surge

Corresponding to the slight price increase of general-purpose and GPU products of Google Cloud in the US region, the reserved prices of similar products in the Azure US region have risen sharply on a 1-year/3-year basis, with an average price increase of 50% for 1-year and 3-year Doubled; interestingly, the on-demand price of the product has not changed at all. In Beijing, the price of general-purpose products has dropped by about 15%.

AWS: up and down

The price of general-purpose products (m4.large) in AWS’s US region was slightly increased by 4%, and the price of its GPU products (p3.2xlarge) in the optimal region (Ningxia, China) was reduced across the board, with the largest price reduction in one year, reaching 30%.

Fifth: In the comparison of GPU products, domestic manufacturers have obvious price advantages.

This year, we uniformly selected NVIDIA Tesla V100 as the base specification for the GPU instance type.

1. The pricing of Google Cloud's GPU products is slightly different from that of the other five cloud vendors. The prices of his GPUs are listed separately, while other vendors are packaged package prices (including CPU/memory/GPU). Therefore, the Google Cloud GPU product we count is the total price that includes the GPU and is equipped with vCPU and memory.

2. The products equipped with NVIDIA Tesla V100 have two specifications: GN10X and GN10Xp. The minimum configurations of the two are 8-core 40G memory and 10-core 40G memory respectively. The CPU of the former is Intel® Xeon® Gold 6133 2.5GHz, and the latter is For Intel® Xeon® Platinum 8255C 2.5GHz. The price of the two specifications on the official website is different on demand, on a monthly basis, on a 1-year basis, and on a 3-year basis. Considering that Tencent singled out the GN10Xp and said that the price of the final order page shall prevail, we chose GN10X as the GPU reference product.

Shanghai Region - Computational CPU:

Shanghai area - general-purpose CPU:

On-demand instance price of general-purpose CPU: Azure is up to 1.25 yuan/hour, and the price gap is quite different.

Reserved instance prices for general-purpose CPUs: monthly/1-year/3-year prices, followed by Alibaba Cloud.

Shanghai area - memory CPU:

On-demand instance prices for memory-based CPUs: Alibaba Cloud has a minimum of 0.66 yuan/hour, and Azure has a maximum of 1.52 yuan/hour.

Reserved instance prices for memory-based CPUs: The monthly prices are lower on Alibaba Cloud. According to the 1-year/3-year price, Alibaba Cloud is in the middle.

Shanghai Region - GPU:

On-demand instance price of GPU: Alibaba Cloud is the highest at 26.46 yuan/hour.

GPU reserved instance price: monthly/1-year/3-year price, Alibaba Cloud is the lowest.

Optimal Region - Computational CPU:

On-demand instance price of computing CPU: Since the 2-core 4G model has been removed from the shelves, we have re-selected the 4-core 8G model to open up a new battlefield. It becomes Azure, which is 1.05 yuan/hour. The price of other manufacturers is between 0.92-1.01 yuan/hour.

Reserved instance price for computing CPU: monthly price. In terms of 1-year/3-year prices, Alibaba Cloud and AWS are both at a low level, and Azure maintains the same high price as last year, with the price in the middle.

Optimal Region - General Purpose CPU:

The on-demand instance price of general-purpose CPU: 0.22 yuan/hour; the prices of other manufacturers vary slightly, basically between 0.65-0.71 yuan/hour.

Reserved instance prices for general-purpose CPUs: On a monthly/1-year/3-year basis, Azure has become the highest price for reserved instances after a round of significant price increases. Alibaba Cloud and AWS are lined up in order between the two.

Optimal area - memory CPU:

On-demand instance prices for memory-based CPUs: Alibaba Cloud is the lowest, at 0.46 yuan/hour; Google Cloud is the highest, at 0.93 yuan/hour.

Reserved instance prices for memory CPUs: The monthly price is the lowest on Alibaba Cloud and the highest on Azure. In terms of 1-year prices, Alibaba Cloud and AWS are relatively low, and Azure is the highest. According to the 3-year price, AWS is the lowest, Alibaba Cloud is the second, and Azure is the highest.

Optimal Region - GPU:

On-demand instance price of GPU: AWS and Azure are tied for the highest price at 21.85 yuan/hour.

GPU reserved instance price: monthly/1-year/3-year price, Alibaba Cloud is the lowest, and Azure is the highest.

Beijing area - computing CPU:

On-Demand Instance Prices for Compute CPUs: Ranking is the same as last year. Azure is next at 1.05 yuan/hour.

Reserved instance price for computing CPU: monthly price, 1-year/3-year price, AWS lowest,

Beijing area - general-purpose CPU:

On-demand instance prices for general-purpose CPUs: Except for Azure, which has reduced prices to a certain extent, the prices of other manufacturers have not changed. AWS's highest price is 1.41 yuan/hour, and the difference between the lowest price and the highest price is relatively large. The price of the remaining three fluctuates between 0.72-0.95 yuan/hour.

Reserved instance price for general-purpose CPU: The general-purpose reserved price in the Beijing area is similar to the on-demand price, and the price is monthly.

Beijing area - memory CPU:

On-demand instance price of memory CPU: Alibaba Cloud has the lowest price at 0.66 yuan/hour. Followed by 0.83 yuan / hour. The prices of other manufacturers fluctuate between 1.22-1.52 yuan/hour.

Reserved instance prices for memory-based CPUs: The monthly prices are lower on Alibaba Cloud. At 1-year price, AWS is the lowest. According to the 3-year price, Alibaba Cloud is in the middle.

Beijing area - GPU:

On-demand instance price of GPU: AWS is the highest at 36.08 yuan/hour.

GPU reserved instance price: monthly/1-year/3-year price, Alibaba Cloud is the lowest. AWS is highest by 1 year.

US Region - Compute CPU:

On-demand instance prices for computing CPUs: The overall prices of the six cloud vendors are relatively stable, with the lowest price being 0.92 yuan/hour, followed by Google Cloud at 1.01 yuan/hour, and AWS the highest at 1.21 yuan/hour.

Reserved instance price for computing CPU: The monthly price is lower than that of Alibaba Cloud. According to the 1-year/3-year price, Alibaba Cloud is the lowest. Both AWS and Azure have higher prices.

US Region - General Purpose CPU:

On-demand instance prices for general-purpose CPUs: There is little price difference between the five cloud vendors. After a relatively obvious price reduction, 0.60 yuan/hour became the local lowest price. After a small price increase, Google Cloud and Alibaba Cloud It is followed by 0.68 yuan/hour, and AWS is the highest at 0.71 yuan/hour.

The price of reserved instances for general-purpose CPUs: the monthly/1-year/3-year price, Alibaba Cloud has the lowest price, ranking next after a wave of price reductions. Azure and AWS are relatively expensive.

US Region - Memory CPU:

On-demand instance prices for memory-based CPUs: There is little price difference between the five cloud vendors, the lowest being 0.79 yuan/hour, and the highest for Google Cloud, 0.93 yuan/hour. The price of other manufacturers is between 0.85-0.90 yuan/hour.

Reserved instance prices for memory CPUs: monthly/yearly prices, Alibaba Cloud is the lowest, and Azure is the highest. According to the 3-year price, Alibaba Cloud is the second, and Azure is the highest.

US Region - GPU:

On-demand instance price of GPU: Alibaba Cloud is 18.26 yuan/hour. AWS and Azure are the highest, both at 21.85 yuan/hour.

GPU reserved instance price: The monthly price is lower for Alibaba Cloud. According to the 1-year/3-year price, Alibaba Cloud is the lowest and Azure is the highest.