While preparing for the exam DP-420 Designing and Implementing Native Applications Using Microsoft Azure Cosmos DB, one significant area to understand is the distinction between serverless and provisioned models. Choosing the right model can drastically impact application performance, cost-effectiveness and overall cloud strategy.
Azure Cosmos DB offers a choice of two models in which you can run your applications: Serverless and Provisioned Throughput. The main differentiator between the two models is how the application data is read and written, and how these operations are billed.
1. Serverless Model
Azure is often associated with the concept of “serverless”. In a serverless model, the computational resources are fully managed by Azure with no dedicated resources. The compute resources scale automatically depending on the demand and the number of requests made. This model is typically billed based on the actual request units (RU) consumed per operation, which effectively eliminates any idle capacity cost. As a result, the serverless model is ideal for applications with unpredictable or sporadic traffic patterns.
Serverless Example:
Using Azure Cosmos DB Serverless, you only pay for the total Request Units (RUs) consumed by database operations and the consumed storage. There are no minimums, no pre-provisioned throughput and no charge for the standby capacity. When the demand subsides, effectively there is no cost.
2. Provisioned Throughput Model (Manual or Autoscale)
In a Provisioned Throughput model, you allocate a certain amount of Resources Units (RUs) for your database or a container. These RUs offer a specific level of performance capacity, which remains constant and is billed hourly.
A. Manual Provisioned Throughput: You decide and set the number of RUs which remain fixed unless manually changed.
B. Autoscale Provisioned Throughput: You set the maximum number of RUs, and Cosmos DB will automatically scale the RUs within a defined range based on the workload.
Provisioned Throughput Example:
Suppose you have an application with a steady workload. In that case, you might want to opt for the Provisioned Throughput model where you pay for a certain amount of Request Units(RUs) per second, offering a predictable performance level, billed irrespective of the usage.
Comparison: Serverless vs. Provisioned Throughput
| Serverless | Provisioned Throughput | |
| Cost | Pay as you Go | Fixed, based on RUs provisioned |
| Scalability | Automatic, on-demand | Manual or Autoscale |
| Ideal For | Sporadic or unpredictable workloads | Steady or predictable workloads |
| Max RU/s per database | 5000 RU/s | Unlimited, based on provision |
While considering the option that’s best for your application, consider the workload patterns, data volume, and operational requirements. A serverless model can help cut costs and manage sporadic traffic while a provisioned model provides consistency and assurance of power availability for heavier or steady workloads.
As you prepare for Microsoft’s DP-420 exam, understanding these models’ nuances will provide more command over designing and implementing native applications using Microsoft Azure Cosmos DB and its different operational models.
Practice Test
True or False: Serverless computing models always cost more than provisioned models.
- True
- False
Answer: False
Explanation: The costs depend on specific use-cases. A serverless model can save costs by only charging for the actual consumption of resources, while provisioned models may cost more due to unused resources.
What advantages does a serverless computing model have over a provisioned model? (Multiple Select)
- A. Automatic scaling
- B. Pay for actual usage
- C. Better predictability of cost
- D. Greater control over instances
Answer: A, B
Explanation: Serverless computing automatically scales based on demand and charges only for actual usage, reducing overhead.
In the context of Azure Cosmos DB, are provisioned throughput models generally best suited for workloads with unpredictable traffic?
- True
- False
Answer: False
Explanation: Provisioned throughput is best for workloads with predictable traffic, as it allows users to allocate and pay for resources in advance.
Which of the following should be considered when choosing between serverless and provisioned models (Multiple Select)?
- A. The predictability of traffic
- B. The scale of operations
- C. The need for cost control
- D. The color of the company logo
Answer: A, B, C
Explanation: Relevant factors are traffic predictability, scale of operations and the need for cost control; the company logo color is irrelevant.
Is it possible to switch between provisioned and serverless models in Azure Cosmos DB?
- Yes
- No
Answer: No
Explanation: As of now, you can’t switch between different models directly in Azure Cosmos DB.
True or False: Serverless models involve more infrastructure management than provisioned models.
- True
- False
Answer: False
Explanation: Serverless models abstract away infrastructure management, allowing developers to focus on code.
What is the billing model for a provisioned throughput in Azure Cosmos DB?
- A. Pay per request
- B. Pay per use
- C. Flat rate
- D. Pay for provisioned resources
Answer: D
Explanation: Provisioned throughput is billed based on the amount of resources provisioned, not actual usage.
In the context of Azure Cosmos DB, can you change provisioned throughput at any time?
- Yes
- No
Answer: Yes
Explanation: Provisioned throughput in Azure Cosmos DB is elastic and can be increased or decreased anytime as per need.
True or False: Serverless models in Azure Cosmos DB are not suitable for high throughput operations.
- True
- False
Answer: True
Explanation: For high throughput or low latency operations, a provisioned model is more suitable as serverless models may have cold start delays.
For a small application with unpredictable and highly variable traffic, which computing model would be more economical?
- Serverless
- Provisioned
Answer: Serverless
Explanation: For small applications with variable traffic, serverless models can be more economical as they only charge for actual consumption.
Which option is best for cases where you want greater control over your resources?
- Serverless
- Provisioned
Answer: Provisioned
Explanation: Provisioned models offer greater control over instances and resources, making them suitable when you need more management control.
True or False: The provisioned model in Azure Cosmos DB allows for automatic and instant scaling.
- True
- False
Answer: False
Explanation: While the provisioned model does allow for scaling, it isn’t automatic and instant like in serverless models.
True or False: Choosing between serverless and provisioned models in Azure Cosmos DB is permanent and cannot be changed later.
- True
- False
Answer: True
Explanation: Once chosen, the model can’t be changed directly, though you can migrate data to a different model’s database.
Which model, serverless or provisioned, allows for more precise budgeting?
- Serverless
- Provisioned
Answer: Provisioned
Explanation: With provisioned models, you pay for what you provision, making costs predictable and budgeting easier.
Does Azure Cosmos DB’s serverless model offer a free tier?
- Yes
- No
Answer: No
Explanation: As of now, Cosmos DB’s serverless model doesn’t include a free tier of usage.
Interview Questions
What is the primary advantage of the serverless model in comparison to the provisioned model?
Serverless computing abstracts away the underlying infrastructure, allowing developers to focus on writing code and not on managing servers, capacity, and availability.
Can you name a scenario where the serverless model is the most appropriate choice?
Serverless model is most appropriate for applications that are unpredictable in terms of traffic, as serverless can easily scale to handle such fluctuations.
What is the main advantage of provisioned throughput model in Azure Cosmos DB?
The main advantage of the provisioned throughput model is that it provides guaranteed high availability, throughput, and low latency at any scale.
When should you choose the provisioned model over serverless in Azure Cosmos DB?
Choose provisioned model for applications that require a high level of performance and predictable millisecond latencies, where the throughput needs are known and fairly consistent.
What are the pricing models for serverless and provisioned throughput in Azure Cosmos DB?
Serverless is billed based on the resources consumed for each operation while provisioned throughput is billed hourly for the throughput that you provision.
Can you highlight the key difference between Serverless and Provisioned models?
The key difference between these two models is how they handle capacity. Serverless scales up and down automatically based on demand, while provisioned allocates a fixed amount of capacity.
In which model can Azure Cosmos DB automatically and instantly scale?
Azure Cosmos DB can automatically and instantly scale with the serverless mode.
Can you highlight any trade-offs of using the serverless model?
Serverless model trade-offs include a potentially longer cold-start time when no executions have occurred for a certain time period.
How does the provisioned model handle spikes in traffic?
With provisioned throughput, you have to manually scale up throughput ahead of time to handle spikes in traffic.
Can you switch between the serverless and provisioned capacity mode?
Yes. Azure Cosmos DB allows you to switch between serverless and provisioned capacity mode but there are specific guidelines to be followed as detailed in the Azure documentation.
What is the major limitation of the serverless model in Cosmos DB?
The major limitation of the serverless model is a potential higher latency compared to provisioned model due to the time required to allocate resources on-demand.
In which model, the throughput remaining in an hour is not carried over to the next hour?
This situation occurs in the provisioned model. Here, the throughput remaining in an hour is not carried over.
What factors should be considered when choosing between serverless and provisioned models?
Factors to consider include the expected throughput, traffic patterns, data storage requirements, and the necessity for predictable performance.
Which Azure Cosmos DB model is more suitable for large scale applications?
The provisioned model is more suitable for large scale applications due to its capability to provide high availability and predictable performance.
Can serverless model in Cosmos DB handle large-scale, heavy load applications?
Serverless model can handle varying workloads but it may not provide the same level of predictable performance as seen with provisioned throughput, especially for heavy-load, large-scale applications.
extutorials.com