Context
When you embrace Azure or another cloud provider, you’ll discover an environment that offers out of the box a high variety of options for each type of action that you need.
Let us take computation as an example. If on-premises you would have the option only on top of VMs, where on top of them you would be able to put any other solution that you would like, when you are on Azure, things are looking different.
Beside the classical VMs, on Microsoft Azure you have: Scale Sets, App Services, Functions, AKS (Azure Container Services), Container Instances, Batch, Service Fabric and other Cloud Services (classical Web and Worker role for example).
Problem
Having so many options, selecting the best service for your needs might be tricky. It is not about taking the wrong decision; it is more about selecting the right service for your needs, taking into account your context and expectations.
This is one of the most challenges steps during the initial phases of the project that can ensure the success of failure of a project.
Internal Fight
When you need to decide what are the Azure (cloud) components that will gone be used inside your solutions, in most of the cases there might be a friction between two groups. One of them is trying to go with greatest and latest services and technologies and the other one is trying to go safe, with a classical approach.
A good example is:
vs
Both parties are proposing good ways on how you can approach the problem. In the end, you need to draw a risk matrix and see what are the pros and cons of each approach.
A hidden risk
Having at a click distance an ecosystem with so many services available, the technical teams are driven to use as many services as possible. On one side, this is good, because using SaaS reduce the responsibility of the team, but using to many services adds also extra complexity.
For each service, you need to know the pricing model, behavior, limitations, API, configuration and many more. This might not sounds too much… or not? Because the team will try to use the latest services and create a state-of-the-art solution, you can end-up with 14-16 services mixed up together for a simple solution that could use just one computation solution and a storage.
Example
Let us imagine that we want to develop a system that is processing documents. A classical approach would be:
Another approach that is good as the first one could contain:
Both solution are correct, there is nothing’s wrong with the first one or with the second one. On the second approach there is an extra complexity layer added by extra services. We need to put this in balancer when we take a look over product requirements and NFRs.
Conclusion
Be careful when you combine too many services from a cloud provider. Each extra service adds extra complexity to the system, making it more complex to manage and develop. There is a thin line between a good solution and over-engineering when you have so many services at a click distance.
When you embrace Azure or another cloud provider, you’ll discover an environment that offers out of the box a high variety of options for each type of action that you need.
Let us take computation as an example. If on-premises you would have the option only on top of VMs, where on top of them you would be able to put any other solution that you would like, when you are on Azure, things are looking different.
Beside the classical VMs, on Microsoft Azure you have: Scale Sets, App Services, Functions, AKS (Azure Container Services), Container Instances, Batch, Service Fabric and other Cloud Services (classical Web and Worker role for example).
Problem
Having so many options, selecting the best service for your needs might be tricky. It is not about taking the wrong decision; it is more about selecting the right service for your needs, taking into account your context and expectations.
This is one of the most challenges steps during the initial phases of the project that can ensure the success of failure of a project.
Internal Fight
When you need to decide what are the Azure (cloud) components that will gone be used inside your solutions, in most of the cases there might be a friction between two groups. One of them is trying to go with greatest and latest services and technologies and the other one is trying to go safe, with a classical approach.
A good example is:
- A solution based on Azure VM Scale Sets, Azure SQL and Azure Storage
vs
- A solution on top of containers (AKS), with Azure Cosmos DB and Azure MySQL
Both parties are proposing good ways on how you can approach the problem. In the end, you need to draw a risk matrix and see what are the pros and cons of each approach.
A hidden risk
Having at a click distance an ecosystem with so many services available, the technical teams are driven to use as many services as possible. On one side, this is good, because using SaaS reduce the responsibility of the team, but using to many services adds also extra complexity.
For each service, you need to know the pricing model, behavior, limitations, API, configuration and many more. This might not sounds too much… or not? Because the team will try to use the latest services and create a state-of-the-art solution, you can end-up with 14-16 services mixed up together for a simple solution that could use just one computation solution and a storage.
Example
Let us imagine that we want to develop a system that is processing documents. A classical approach would be:
- Azure Storage
- Azure SQL
- Azure Service Bus
- Azure App Services
- Azure Batch
Another approach that is good as the first one could contain:
- Azure Storage
- Azure SQL
- Azure Cosmos DB
- Azure App Services
- Azure Container Services
- Azure Functions
- Azure Data Factory
- Azure Logic Apps
Both solution are correct, there is nothing’s wrong with the first one or with the second one. On the second approach there is an extra complexity layer added by extra services. We need to put this in balancer when we take a look over product requirements and NFRs.
Conclusion
Be careful when you combine too many services from a cloud provider. Each extra service adds extra complexity to the system, making it more complex to manage and develop. There is a thin line between a good solution and over-engineering when you have so many services at a click distance.
Comments
Post a Comment