The Top Clouds Evaluated Such That You Don’t Need To Repeat Our Mistakes | by Ilya Yalchyk | May, 2022

Opinion on popular cloud service providers from an ML engineer

Are you struggling to make sense of different cloud service providers? Are your cloud costs exploding? Are you afraid to make a strategic mistake when switching providers? In this article, we have gathered all of our knowledge and experiences to help you out.

Clouds have been growing and being extended for years, and since then they have become a crucial part of many software systems. There are over 17k open positions for cloud engineers in Germany on LinkedIn. For many companies, there is no question on whether they should use a cloud or not. For example, I work in a startup company, and for us, the possibility to use existing robust and scalable infrastructure with servers and analytical software significantly outweighs the costs and enables us to focus more on the creative side of the software rather than on technical server issues. A more relevant question for us is which cloud to choose, and here I want to share my thoughts on it.

In this article, I will briefly analyze three major clouds — AWS, Azure, and GCP. Along with some objective characteristics, I will share my personal experience working with each of them, including the problems I faced and how I solved them.

Modern clouds offer hundreds of services, and of course, we cannot analyze all of them. Though, some of them are used more often than others. For example, what we usually need from the clouds in our projects is the following:

· An SQL or file-based data storage.

· A batch service to run our ETL processes on HTTP requests or time triggers.

· A lightweight API service to handle quick requests to our models.

Since we are focusing on developing lightweight machine learning models, we don’t consider the availability of high-performant machines. What we usually need are inexpensive virtual machines or preemptible cost-efficient clusters.

Google Cloud Platform was the first cloud I met, so let’s start with it first.

Among the 3 giants, GCP is the least popular choice. Nevertheless, its popularity grows relatively fast. For the last two years, it gained almost 2% of the market share growing from 7.2% to 9.1% [1]. Its lower market share is explainable by the fact that GCP was launched not so long time ago (one of its core services — App Engine — became generally available in 2011) [2].

What does it mean for a cloud service provider to be young? In the case of GCP, it results in a lesser number of services available. While AWS and Azure offer over 200 services, GCP has slightly over 100. For example, you won’t find dedicated services for blockchain development, quantum computing, and graph databases in GCP now. Although it could be a deal-breaking drawback for GCP if your business is built around such technologies, it could be even an advantage in more traditional cases — your team won’t get lost in the infinite lists of configurations and services.

Historically, people have been criticizing GCP for a lower number of data centers available. Indeed, it is an important thing to consider. It might have appeared that AWS and Azure had their data centers closer to you or your clients, while GCP data centers were further, which would have increased the latency. Fortunately, this drawback is not relevant anymore since Google opened multiple new regions and zones and keeps adding new ones. In terms of the number of regions and zones, Google outperforms AWS — a former leader in this context.

As a relatively young competitor in the cloud computing market, GCP can also offer better deals in terms of price. For example, GCP offers the cheapest virtual machine “e2-micro-preemptible” with 2 vCPU and 1 GB memory. Its price is 48% lower than “t4g.nano” from AWS and 5 times lower than “A0” from Azure. If you need a cheap dedicated PostgreSQL server, GCP also can offer one with a price 25% lower than the competitors [2]. In general, GCP is cheaper on the lower-end level. More performant instances usually cost approximately the same as their analogs from other cloud providers. It’s important to know that GCP doesn’t have cheap memory-optimized instances — there are only high performant and extremely expensive ones with 40 vCPUs.

Personal experience with GCP

What I really like about GCP is the web interface and the documentation, which are implemented in a very concise and user-friendly manner. Personally, I think it’s one of the main advantages directly affecting the amount of time your team might need to spend to implement a solution.

GCP seemed more stable than Azure, but still not as stable as AWS. For example, we faced one annoying bug in App Engine. Every request to 3rd party services took at least 2 minutes to execute from App Engine, while it needed at most 100 ms from my local environment. All we needed to do is to switch from a “Standard” to a “Flex” environment. Interestingly, Java components worked fine with the “Standard” environments, so the issue was related to the Google Cloud SDK for Python only.

App Engine surprised us one more time. One day, we reviewed the costs forecast and realized that App Engine was generating significant costs five times higher than our expectations. For some reason, five App Engine instances were assigned to some staled versions, which stayed there after re-deployments, and they were generating costs despite being inactive.

To conclude, lower prices for inexpensive machines, a lesser number of configuration options and service alternatives make GCP a perfect choice for smaller companies. Some might consider another advantage advertised by Google, which is hard to check though. They declare that Google Cloud data centers run on half the energy of a typical data center and run on 100% renewable energy where available [3].

GCP Tips

GCP is suitable for you if:

• You are a start-up company.
• You can’t invest much time in learning AWS and dealing with Azure bugs.
• You don’t need much flexibility and configuration facilities from the cloud.
• You are ready to accept the approaches dictated by the platform.
• You need either a general-purpose or a compute-optimized solution, but not a memory-optimized one.

If you decide to go with GCP, here are my recommendations for developers:

• Don’t use App Engine Standard environments — big brother G wants you to use rather Flex environments, otherwise, they’ll punish you.
• Review cost analysis regularly to make sure there are no surprising costs.
• Make sure you clean up redundant App Engine application versions to prevent G from robbing you.

Currently, AWS is the market leader in cloud services with a market share of 49.2%, although slowly losing its position to Azure and GCP [1]. It’s also the oldest cloud service provider among our three competitors since it was launched in 2002. Honestly, I never worked with AWS intensively apart from running simple EC2 instances, but we always considered AWS as an alternative cloud service provider for our projects.

AWS has been well-known as an industry standard for large enterprise projects. For this reason, AWS provides the largest choice of machines including extremely performant ones with 448 CPUs, 12 TB of memory, and a 100 Gigabit network for $132 per hour. Apart from renting out virtual machines, AWS provides over 200 services for different needs. Comparison of specific services between AWS, Azure, and GCP is a topic for another article, but in general AWS pricing is averagely higher, although when it comes to the cheapest configurations, AWS is in-between GCP and Azure in terms of price [4].

An obvious advantage of AWS is its maturity. AWS does not change a lot. You hardly find any bugs in the most popular services, the documentation is broad and well-written. Its high popularity means that you have higher chances to find new employees who are familiar with AWS as well as clients who also work with AWS.

Another advertised advantage of AWS is its flexibility. Indeed, you can satisfy any project need in terms of a cloud configuration, but the amount of possible configuration needed naturally increases complexity and time to configure. If your team doesn’t have experts in AWS, the complexity can get overwhelming sometimes.

Personal experience with AWS

AWS didn’t bring any surprises, but two things seemed inconvenient to me. The first one — if you create a tag policy, it is available only within the current region. When I switched the region, I had to re-create the policy, although there could be an option to make the policy global.

The second inconvenience was also related to the tag policies. When I created an enforced tag policy, AWS didn’t let me create EC2 instances without those tags (which is what I wanted), but I still had to type in the tag names and find them in the dropdown. I wanted to see the tags already assigned with empty values and prompting to fill in the values before the instance creation. It appeared that I needed to implement a Service Control Policy under the AWS Organization tool to enable such behavior. I found this procedure quite complicated for such a simple task.

AWS Tips

AWS is suitable for you if:

• You are rich.
• You have AWS experts in your team.
• You build an enterprise-level long-term project.
• OR you just want to rent a cheap virtual machine, and you don’t care about all the other facilities.

If you decide to go with AWS, here are my recommendations for developers:

• Preliminarily learn main concepts from some external materials. I highly recommend “AWS Fundamentals Specialization” on Coursera — not everything on AWS is clear by intuition.
• Dedicate some time to go through the documentation and understand how the documentation is organized.

For multiple reasons, some of our clients preferred Microsoft Azure as a cloud service provider. I notice that it is especially popular in Germany. One of the reasons could be due to the popularity of Microsoft Teams or Microsoft Outlook, which operate under Microsoft 365 account and provide access to all other Microsoft tools and services including Azure. On the market, Azure is between GCP and AWS with a market share of 33.1%, and the data shows that its popularity grows [1]. When it comes to the number of services, Azure is also considered as a compromise choice between AWS and GCP. Apart from that, Microsoft advertises deep integration of Azure with other Microsoft products.

Although Azure is considered to be a compromise solution in terms of price, it is super expensive on the lower-end. Their cheapest virtual machine of general purpose “A0” costs 5 times more and has less memory than an analog from GCP. It’s also 3.5 times more expensive than an analog from AWS. The cheapest PostgreSQL instance “B1MS” is also more expensive than its analogs from GCP and AWS, although the difference is not that drastic here [5]. For more performant types of configuration, Azure servers cost the same or slightly less than AWS. One more thing to consider is that Windows machines are usually cheaper on Azure than on AWS or GCP. [6]

People often report issues with the Azure documentation, including inconsistencies, mistakes, and missing parts. The same concerns the Python SDK. These are just rumors, you might say, but here is what I experienced with Azure.

Personal experience with Azure

Azure gave me the worst impression in terms of its user experience. For example, what do you think the “No configurations available, your region is under maintenance” error means when you are choosing a database configuration? Would you decide to wait a little while the maintenance is over? Ha, you will wait forever! In Azure, this error means that the corresponding resource provider registration is missing in the subscription settings.

If you were a developer of the Azure portal, what would you show first to a user who opens the “Cost analysis” panel? If your answer is something like “estimated costs for the next month”, then you are far behind. Microsoft is showing a “Bad Request” error with lots of details to give you a chance to think — which type of forecast you want to see — prepare morally, and learn something about Azure internals meanwhile.

Have you discovered Azure Machine Learning? It’s such a great platform! If you are creating real-time endpoints, you know that your source code will be deliberately delivered to your service, and you can use your modules in your scoring functions. This is very convenient. The platform also allows you to create a batch endpoint. Why would you expect that your source code is delivered either? If you are a unique person who needs their modules in their batch processes, think of your own unique way to deliver the code there! Are you desperate and want to find the answer in the Azure documentation? Well, we all were young and naive one day. Ah yes, you can’t even deploy batch endpoints using Azure ML SDK for Python. Either create a published pipeline or use Azure CLI, or use Azure ML Portal, because a convenient deployment is a privilege only for real-time endpoints.

You can use Azure Machine Learning also to punish junior developers or reduce salary expectations for interviewees in your company. For example, just add the Flask module to the conda dependencies list and ask them to deploy this service. They will have a great time investigating a very interesting error. Azure also guarantees that they will not find any hints in the documentation.

Azure Machine Learning is not a bad platform. If you go through all the inconsistencies, then it works fine.

3.2. Azure — Tips

Microsoft Azure is suitable for you if:

• You use the Microsoft Office stack (Word, Teams, OneDrive, SharePoint, etc.) and/or C# programming language.
• You head neither for the cheapest servers nor for the most expensive ones — you need something in the middle.
• You need a memory-optimized solution rather than a general-purpose or a compute-optimized one.
• You read about the current bugs and inconsistencies in Azure, and it does not scare you.

If you decide to go with Microsoft Azure, here are my recommendations for developers:

• If you can’t create a service, because Azure servers are under maintenance for more than a couple of minutes — check out your permissions and registrations under the “Resource providers” panel.
• If you see any strange errors on the Azure Portal — just change the filters’ values.
• If you use Azure Machine Learning, and your scoring function cannot locate your source code — deliver the code as a Model and add it explicitly to the sys.path in the init function.
• If you use Azure Machine Learning, don’t use Batch Endpoints — it looks like they are not ready yet — just use the regular Published Pipelines. In fact, “Batch endpoint” is just a wrapper around a published pipeline.
• Don’t include flask in your Azure conda environment specification.

I mentioned multiple differences that can help you to choose the right cloud service. At the same time, I must admit, that all the mentioned companies provide approximately the same services for approximately the same price. In this context, I would rather rely on my team’s expertise more than on specific features of a certain cloud service provider. From my experience, developers’ time costs more than you can save from choosing a better cloud service. Thus, if my team comprises experts in Azure, I would stick to Azure despite having a bad experience with it.

I am amazed how much these cloud service providers simplify work by taking over so many things to care about. Sad truth is that this simplification costs not only money but also our time to learn them deeper and tackle all upcoming challenges. Nevertheless, I am happy that my cloud journey continues and my expertise in clouds grows day over day.

[1] Vailshery, L. S. (2022, February 16). Hyperscaler Cloud Market Market Share 2022. Statista. Retrieved May 6, 2022, from https://www.statista.com/statistics/1202770/hyperscaler-iaas-paas-market-share

[2] Google Cloud Pricing Calculator. (n.d.). Google Cloud. Retrieved May 6, 2022, from https://cloud.google.com/products/calculator/#id=1abd7a80-bf53-4f77-8006-db48d924dd9c

[3] What makes Google Cloud Platform different? (n.d.). Google Cloud. https://cloud.google.com/free/docs/what-makes-google-cloud-platform-different‌

[4] “AWS Pricing Calculator,” calculator.aws. https://calculator.aws/#/estimate?id=171ef16e56c643cb303bd94f950af0c0343058ad (accessed May 06, 2022).‌

[5] Cloud Computing Services | Microsoft Azure. (n.d.). Azure.microsoft.com. Retrieved May 6, 2022, from https://azure.com/e/749b0305ecd74af280c0ab84ad998d8e

[6] Santis, P. (2021, August 26). Ultimate Cloud Pricing Comparison: AWS vs. Azure vs. Google Cloud in 2021. CAST AI. https://cast.ai/blog/ultimate-cloud-pricing-comparison-aws-vs-azure-vs-google-cloud-in-2021/

Opinion on popular cloud service providers from an ML engineer

Are you struggling to make sense of different cloud service providers? Are your cloud costs exploding? Are you afraid to make a strategic mistake when switching providers? In this article, we have gathered all of our knowledge and experiences to help you out.

Clouds have been growing and being extended for years, and since then they have become a crucial part of many software systems. There are over 17k open positions for cloud engineers in Germany on LinkedIn. For many companies, there is no question on whether they should use a cloud or not. For example, I work in a startup company, and for us, the possibility to use existing robust and scalable infrastructure with servers and analytical software significantly outweighs the costs and enables us to focus more on the creative side of the software rather than on technical server issues. A more relevant question for us is which cloud to choose, and here I want to share my thoughts on it.

In this article, I will briefly analyze three major clouds — AWS, Azure, and GCP. Along with some objective characteristics, I will share my personal experience working with each of them, including the problems I faced and how I solved them.

Modern clouds offer hundreds of services, and of course, we cannot analyze all of them. Though, some of them are used more often than others. For example, what we usually need from the clouds in our projects is the following:

· An SQL or file-based data storage.

· A batch service to run our ETL processes on HTTP requests or time triggers.

· A lightweight API service to handle quick requests to our models.

Since we are focusing on developing lightweight machine learning models, we don’t consider the availability of high-performant machines. What we usually need are inexpensive virtual machines or preemptible cost-efficient clusters.

Google Cloud Platform was the first cloud I met, so let’s start with it first.

Among the 3 giants, GCP is the least popular choice. Nevertheless, its popularity grows relatively fast. For the last two years, it gained almost 2% of the market share growing from 7.2% to 9.1% [1]. Its lower market share is explainable by the fact that GCP was launched not so long time ago (one of its core services — App Engine — became generally available in 2011) [2].

What does it mean for a cloud service provider to be young? In the case of GCP, it results in a lesser number of services available. While AWS and Azure offer over 200 services, GCP has slightly over 100. For example, you won’t find dedicated services for blockchain development, quantum computing, and graph databases in GCP now. Although it could be a deal-breaking drawback for GCP if your business is built around such technologies, it could be even an advantage in more traditional cases — your team won’t get lost in the infinite lists of configurations and services.

Historically, people have been criticizing GCP for a lower number of data centers available. Indeed, it is an important thing to consider. It might have appeared that AWS and Azure had their data centers closer to you or your clients, while GCP data centers were further, which would have increased the latency. Fortunately, this drawback is not relevant anymore since Google opened multiple new regions and zones and keeps adding new ones. In terms of the number of regions and zones, Google outperforms AWS — a former leader in this context.

As a relatively young competitor in the cloud computing market, GCP can also offer better deals in terms of price. For example, GCP offers the cheapest virtual machine “e2-micro-preemptible” with 2 vCPU and 1 GB memory. Its price is 48% lower than “t4g.nano” from AWS and 5 times lower than “A0” from Azure. If you need a cheap dedicated PostgreSQL server, GCP also can offer one with a price 25% lower than the competitors [2]. In general, GCP is cheaper on the lower-end level. More performant instances usually cost approximately the same as their analogs from other cloud providers. It’s important to know that GCP doesn’t have cheap memory-optimized instances — there are only high performant and extremely expensive ones with 40 vCPUs.

Personal experience with GCP

What I really like about GCP is the web interface and the documentation, which are implemented in a very concise and user-friendly manner. Personally, I think it’s one of the main advantages directly affecting the amount of time your team might need to spend to implement a solution.

GCP seemed more stable than Azure, but still not as stable as AWS. For example, we faced one annoying bug in App Engine. Every request to 3rd party services took at least 2 minutes to execute from App Engine, while it needed at most 100 ms from my local environment. All we needed to do is to switch from a “Standard” to a “Flex” environment. Interestingly, Java components worked fine with the “Standard” environments, so the issue was related to the Google Cloud SDK for Python only.

App Engine surprised us one more time. One day, we reviewed the costs forecast and realized that App Engine was generating significant costs five times higher than our expectations. For some reason, five App Engine instances were assigned to some staled versions, which stayed there after re-deployments, and they were generating costs despite being inactive.

To conclude, lower prices for inexpensive machines, a lesser number of configuration options and service alternatives make GCP a perfect choice for smaller companies. Some might consider another advantage advertised by Google, which is hard to check though. They declare that Google Cloud data centers run on half the energy of a typical data center and run on 100% renewable energy where available [3].

GCP Tips

GCP is suitable for you if:

• You are a start-up company.
• You can’t invest much time in learning AWS and dealing with Azure bugs.
• You don’t need much flexibility and configuration facilities from the cloud.
• You are ready to accept the approaches dictated by the platform.
• You need either a general-purpose or a compute-optimized solution, but not a memory-optimized one.

If you decide to go with GCP, here are my recommendations for developers:

• Don’t use App Engine Standard environments — big brother G wants you to use rather Flex environments, otherwise, they’ll punish you.
• Review cost analysis regularly to make sure there are no surprising costs.
• Make sure you clean up redundant App Engine application versions to prevent G from robbing you.

Currently, AWS is the market leader in cloud services with a market share of 49.2%, although slowly losing its position to Azure and GCP [1]. It’s also the oldest cloud service provider among our three competitors since it was launched in 2002. Honestly, I never worked with AWS intensively apart from running simple EC2 instances, but we always considered AWS as an alternative cloud service provider for our projects.

AWS has been well-known as an industry standard for large enterprise projects. For this reason, AWS provides the largest choice of machines including extremely performant ones with 448 CPUs, 12 TB of memory, and a 100 Gigabit network for $132 per hour. Apart from renting out virtual machines, AWS provides over 200 services for different needs. Comparison of specific services between AWS, Azure, and GCP is a topic for another article, but in general AWS pricing is averagely higher, although when it comes to the cheapest configurations, AWS is in-between GCP and Azure in terms of price [4].

An obvious advantage of AWS is its maturity. AWS does not change a lot. You hardly find any bugs in the most popular services, the documentation is broad and well-written. Its high popularity means that you have higher chances to find new employees who are familiar with AWS as well as clients who also work with AWS.

Another advertised advantage of AWS is its flexibility. Indeed, you can satisfy any project need in terms of a cloud configuration, but the amount of possible configuration needed naturally increases complexity and time to configure. If your team doesn’t have experts in AWS, the complexity can get overwhelming sometimes.

Personal experience with AWS

AWS didn’t bring any surprises, but two things seemed inconvenient to me. The first one — if you create a tag policy, it is available only within the current region. When I switched the region, I had to re-create the policy, although there could be an option to make the policy global.

The second inconvenience was also related to the tag policies. When I created an enforced tag policy, AWS didn’t let me create EC2 instances without those tags (which is what I wanted), but I still had to type in the tag names and find them in the dropdown. I wanted to see the tags already assigned with empty values and prompting to fill in the values before the instance creation. It appeared that I needed to implement a Service Control Policy under the AWS Organization tool to enable such behavior. I found this procedure quite complicated for such a simple task.

AWS Tips

AWS is suitable for you if:

• You are rich.
• You have AWS experts in your team.
• You build an enterprise-level long-term project.
• OR you just want to rent a cheap virtual machine, and you don’t care about all the other facilities.

If you decide to go with AWS, here are my recommendations for developers:

• Preliminarily learn main concepts from some external materials. I highly recommend “AWS Fundamentals Specialization” on Coursera — not everything on AWS is clear by intuition.
• Dedicate some time to go through the documentation and understand how the documentation is organized.

For multiple reasons, some of our clients preferred Microsoft Azure as a cloud service provider. I notice that it is especially popular in Germany. One of the reasons could be due to the popularity of Microsoft Teams or Microsoft Outlook, which operate under Microsoft 365 account and provide access to all other Microsoft tools and services including Azure. On the market, Azure is between GCP and AWS with a market share of 33.1%, and the data shows that its popularity grows [1]. When it comes to the number of services, Azure is also considered as a compromise choice between AWS and GCP. Apart from that, Microsoft advertises deep integration of Azure with other Microsoft products.

Although Azure is considered to be a compromise solution in terms of price, it is super expensive on the lower-end. Their cheapest virtual machine of general purpose “A0” costs 5 times more and has less memory than an analog from GCP. It’s also 3.5 times more expensive than an analog from AWS. The cheapest PostgreSQL instance “B1MS” is also more expensive than its analogs from GCP and AWS, although the difference is not that drastic here [5]. For more performant types of configuration, Azure servers cost the same or slightly less than AWS. One more thing to consider is that Windows machines are usually cheaper on Azure than on AWS or GCP. [6]

People often report issues with the Azure documentation, including inconsistencies, mistakes, and missing parts. The same concerns the Python SDK. These are just rumors, you might say, but here is what I experienced with Azure.

Personal experience with Azure

Azure gave me the worst impression in terms of its user experience. For example, what do you think the “No configurations available, your region is under maintenance” error means when you are choosing a database configuration? Would you decide to wait a little while the maintenance is over? Ha, you will wait forever! In Azure, this error means that the corresponding resource provider registration is missing in the subscription settings.

If you were a developer of the Azure portal, what would you show first to a user who opens the “Cost analysis” panel? If your answer is something like “estimated costs for the next month”, then you are far behind. Microsoft is showing a “Bad Request” error with lots of details to give you a chance to think — which type of forecast you want to see — prepare morally, and learn something about Azure internals meanwhile.

Have you discovered Azure Machine Learning? It’s such a great platform! If you are creating real-time endpoints, you know that your source code will be deliberately delivered to your service, and you can use your modules in your scoring functions. This is very convenient. The platform also allows you to create a batch endpoint. Why would you expect that your source code is delivered either? If you are a unique person who needs their modules in their batch processes, think of your own unique way to deliver the code there! Are you desperate and want to find the answer in the Azure documentation? Well, we all were young and naive one day. Ah yes, you can’t even deploy batch endpoints using Azure ML SDK for Python. Either create a published pipeline or use Azure CLI, or use Azure ML Portal, because a convenient deployment is a privilege only for real-time endpoints.

You can use Azure Machine Learning also to punish junior developers or reduce salary expectations for interviewees in your company. For example, just add the Flask module to the conda dependencies list and ask them to deploy this service. They will have a great time investigating a very interesting error. Azure also guarantees that they will not find any hints in the documentation.

Azure Machine Learning is not a bad platform. If you go through all the inconsistencies, then it works fine.

3.2. Azure — Tips

Microsoft Azure is suitable for you if:

• You use the Microsoft Office stack (Word, Teams, OneDrive, SharePoint, etc.) and/or C# programming language.
• You head neither for the cheapest servers nor for the most expensive ones — you need something in the middle.
• You need a memory-optimized solution rather than a general-purpose or a compute-optimized one.
• You read about the current bugs and inconsistencies in Azure, and it does not scare you.

If you decide to go with Microsoft Azure, here are my recommendations for developers:

• If you can’t create a service, because Azure servers are under maintenance for more than a couple of minutes — check out your permissions and registrations under the “Resource providers” panel.
• If you see any strange errors on the Azure Portal — just change the filters’ values.
• If you use Azure Machine Learning, and your scoring function cannot locate your source code — deliver the code as a Model and add it explicitly to the sys.path in the init function.
• If you use Azure Machine Learning, don’t use Batch Endpoints — it looks like they are not ready yet — just use the regular Published Pipelines. In fact, “Batch endpoint” is just a wrapper around a published pipeline.
• Don’t include flask in your Azure conda environment specification.

I mentioned multiple differences that can help you to choose the right cloud service. At the same time, I must admit, that all the mentioned companies provide approximately the same services for approximately the same price. In this context, I would rather rely on my team’s expertise more than on specific features of a certain cloud service provider. From my experience, developers’ time costs more than you can save from choosing a better cloud service. Thus, if my team comprises experts in Azure, I would stick to Azure despite having a bad experience with it.

I am amazed how much these cloud service providers simplify work by taking over so many things to care about. Sad truth is that this simplification costs not only money but also our time to learn them deeper and tackle all upcoming challenges. Nevertheless, I am happy that my cloud journey continues and my expertise in clouds grows day over day.

[1] Vailshery, L. S. (2022, February 16). Hyperscaler Cloud Market Market Share 2022. Statista. Retrieved May 6, 2022, from https://www.statista.com/statistics/1202770/hyperscaler-iaas-paas-market-share

[2] Google Cloud Pricing Calculator. (n.d.). Google Cloud. Retrieved May 6, 2022, from https://cloud.google.com/products/calculator/#id=1abd7a80-bf53-4f77-8006-db48d924dd9c

[3] What makes Google Cloud Platform different? (n.d.). Google Cloud. https://cloud.google.com/free/docs/what-makes-google-cloud-platform-different‌

[4] “AWS Pricing Calculator,” calculator.aws. https://calculator.aws/#/estimate?id=171ef16e56c643cb303bd94f950af0c0343058ad (accessed May 06, 2022).‌

[5] Cloud Computing Services | Microsoft Azure. (n.d.). Azure.microsoft.com. Retrieved May 6, 2022, from https://azure.com/e/749b0305ecd74af280c0ab84ad998d8e

[6] Santis, P. (2021, August 26). Ultimate Cloud Pricing Comparison: AWS vs. Azure vs. Google Cloud in 2021. CAST AI. https://cast.ai/blog/ultimate-cloud-pricing-comparison-aws-vs-azure-vs-google-cloud-in-2021/