In this post we will talk about what is a scale unit and what are the benefits of scale unit concept when we are working with a system that is running in a cloud environment.
What is a scale unit?
We can see a scale unit as a group of resources that are grouped together to serve a specific number of clients or requests. This scale unit has a ‘common’ configuration that specifies the resources that are needed by a scale unit.
Let’s assume that we have a scale unit that contains:
Having all of them grouped together we can test the environment at a specific scale. Otherwise we could try to scale our system infinitely, but all of us knows that this is not possible. All the resources under the same scale unit work together for the same purpose.
Each scale unit serve a specific number of clients (or resources). Because the scale unit is fixed we can know exactly what is the throughput of our scale unit - number of requests per second, number of messages that can be consumed, number of access at storage and so on.
In the end we will know exactly the number of clients that we can server or manage for each scale unit and the cost of a scale unit.
Scaling can be made easily without affecting the performance of the system, by adding new units each time.
This means that we scale very easily, by adding new scale units. For each scale unit we know exactly what are the costs. In this way we can estimate the cost easily.
The hardest thing is to separate all each scale units. Between each scale unit we should not have any kind of communication or a central node (a master one). This is the hardest thing to accomplish, because large systems are very complex with a lot of dependencies.
I think that scale unit can help us to be able to predict the necessary request and to scale in a safe way.
In the above example we can see to instances of our scale unit. Each scale unit is mapped to a specific scale unit. There is no communication between scale unit. Each scale unit can be hosted in the same data centers or in different data centers, based on our needs.
In the future, with the new portal. we will be able very easily to create the provisioning for a scale unit and control the provisioning with a few clicks. Azure V2 will allow us to define a JSON file that can be used to provision all the components from our scale unit and connected between them, without having to specify the storage account name and key to the worker roles that need this information (we will be able to do this using a script).
In the next post we will try to see how we map a system that requires 'some' communication between scale units.
What is a scale unit?
We can see a scale unit as a group of resources that are grouped together to serve a specific number of clients or requests. This scale unit has a ‘common’ configuration that specifies the resources that are needed by a scale unit.
Let’s assume that we have a scale unit that contains:
- 2 Azure SQL
- 4 Service Bus Namespaces (with 100 Queues per namespace)
- 8 Worker Roles
- 3 Web Roles
- 2 Different storage accounts
Having all of them grouped together we can test the environment at a specific scale. Otherwise we could try to scale our system infinitely, but all of us knows that this is not possible. All the resources under the same scale unit work together for the same purpose.
Each scale unit serve a specific number of clients (or resources). Because the scale unit is fixed we can know exactly what is the throughput of our scale unit - number of requests per second, number of messages that can be consumed, number of access at storage and so on.
In the end we will know exactly the number of clients that we can server or manage for each scale unit and the cost of a scale unit.
Scaling can be made easily without affecting the performance of the system, by adding new units each time.
This means that we scale very easily, by adding new scale units. For each scale unit we know exactly what are the costs. In this way we can estimate the cost easily.
The hardest thing is to separate all each scale units. Between each scale unit we should not have any kind of communication or a central node (a master one). This is the hardest thing to accomplish, because large systems are very complex with a lot of dependencies.
I think that scale unit can help us to be able to predict the necessary request and to scale in a safe way.
In the above example we can see to instances of our scale unit. Each scale unit is mapped to a specific scale unit. There is no communication between scale unit. Each scale unit can be hosted in the same data centers or in different data centers, based on our needs.
In the future, with the new portal. we will be able very easily to create the provisioning for a scale unit and control the provisioning with a few clicks. Azure V2 will allow us to define a JSON file that can be used to provision all the components from our scale unit and connected between them, without having to specify the storage account name and key to the worker roles that need this information (we will be able to do this using a script).
In the next post we will try to see how we map a system that requires 'some' communication between scale units.
Comments
Post a Comment