Migrating On-Premises Windows HPC Cluster to Azure Batch for Financial Risk Modelling
Question
Your company has an on-premises Windows HPC cluster. The cluster runs an intrinsically parallel, compute-intensive workload that performs financial risk modelling.
You plan to migrate the workload to Azure Batch.
You need to design a solution that will support the workload. The solution must meet the following requirements:
-> Support the large-scale parallel execution of Azure Batch jobs.
-> Minimize cost.
What should you include in the solution?
Answers
Explanations
Click on the arrows to vote for the correct answer
A. B. C. D.B
https://docs.microsoft.com/en-us/azure/batch/batch-technical-overviewTo support the large-scale parallel execution of Azure Batch jobs and minimize cost, the most appropriate option would be to use low-priority virtual machines.
Low-priority virtual machines are a cost-effective option for running batch jobs as they are available at a significantly lower price compared to standard virtual machines. They are best suited for workloads that can tolerate interruptions and do not require high availability. These virtual machines are offered on the surplus capacity of Azure and can be reclaimed by Azure if required by other customers, with a short grace period.
Azure Batch allows for the deployment and management of large-scale parallel and high-performance computing (HPC) workloads in Azure. The parallel execution of batch jobs can be achieved by dividing the workload into smaller tasks and assigning them to individual virtual machines for processing. As the workload can be divided into smaller tasks, the use of low-priority virtual machines for parallel processing helps reduce the cost of the overall solution.
Using basic A-series virtual machines or burstable virtual machines may not be ideal for compute-intensive workloads that require high performance as these machines may not provide the necessary processing power. Additionally, using Azure virtual machine sizes that support the Message Passing Interface (MPI) API may not be cost-effective for the given workload as they are typically designed for specialized workloads that require high-performance inter-node communication.
In conclusion, to support the large-scale parallel execution of Azure Batch jobs and minimize cost, the best option would be to use low-priority virtual machines.
