AZ-305 Question 171
Select 2You are designing a highly available microservices architecture in Azure to process and store large volumes of IoT data. The solution will use Azure Kubernetes Service (AKS) for the microservices, Azure Event Hubs to ingest real-time data, and Azure Cosmos DB for storage. The organization requires minimal downtime in case of a regional outage and wants to handle traffic spikes globally. Which two actions should you recommend to fulfill these requirements?
- A
Deploy AKS clusters in an active-active setup across multiple regions using Azure Front Door for load balancing.
- B
Configure Azure Cosmos DB with multiple write regions.
- C
Choose Basic tier for Azure Event Hubs to simplify deployment.
- D
Use a single-region deployment for AKS to reduce complexity.
- E
Store persistent data on local ephemeral disks attached to the AKS nodes.
Show answer and explanation
Correct answers: A, B
Explanation
To design an architecture that is both high-throughput and highly available, you should deploy services in multiple regions and leverage built-in geo-redundancy. Azure Front Door provides layer 7 load balancing and global routing, while Azure Cosmos DB’s multi-region write capability ensures minimal disruption and data loss if one region goes offline. For more guidance, refer to the official Azure best practices documentation: https://docs.microsoft.com/azure/architecture/best-practices/ and https://docs.microsoft.com/azure/cosmos-db.
- A. Correct.
Deploying AKS clusters in multiple regions with Azure Front Door helps ensure immediate failover and continuous availability if one region experiences an outage. This approach also supports global traffic distribution.
- B. Correct.
Enabling multiple write regions in Azure Cosmos DB provides geo-redundancy and near real-time replication, minimizing downtime and data loss during regional failures.
- C. Incorrect.
Using Basic tier for Azure Event Hubs saves cost but lacks important features such as geo-disaster recovery and higher throughput options, which are critical for highly available solutions handling large data volumes.
- D. Incorrect.
A single-region deployment cannot meet the requirement for minimal downtime in the event of a regional outage. Multi-region deployment is necessary for high availability.
- E. Incorrect.
Relying on ephemeral disks for storing stateful data is risky. Ephemeral storage does not persist data if the node is restarted or replaced, which could lead to data loss.