You have a batch inference model, size of around 2 GB.
You need to deploy this model for batch inferencing tasks and you need to choose a compute for production use.
Which compute should you choose?
Click on the arrows to vote for the correct answer
A. B. C. D.Answer: B.
Option A is incorrect because ACI is recommended for development and testing purposes and it is suitable only for models less than 1 GB in size.
Option B is CORRECT because Azure compute cluster is a cost-effective way to run experiments that need to handle large volumes of data.
It is an option that can provide a containerized environment for the deployed models and it is recommended for service endpoints that need to periodically process batches of data.
Option C is incorrect because HDInsight is a popular, Apache Spark based big-data platform which can be used as an attached compute for training models.
It is not a compute target for inference inference scenarios.
Option D is incorrect because local computes are suitable for development and testing, but they are not appropriate resources for batch inferencing on the basis of large amounts of data.
Reference:
When it comes to deploying a batch inference model for production use, several factors need to be taken into consideration, such as the size of the model, the frequency of the inference tasks, and the resources needed to perform the inference tasks efficiently. In this scenario, the model size is around 2 GB, and we need to choose a compute for production use.
Azure offers several options for deploying batch inference models, including Azure Container Instances, Azure ML Compute Cluster, HDInsight, and Local compute.
Azure Container Instances (A) are a service that allows you to run containers on Azure without the need to manage virtual machines or clusters. ACI is suitable for scenarios where you need to run a single container or a small number of containers. However, it may not be the best option for running large batch inference models since it has a limited amount of memory and CPU resources.
Azure ML Compute Cluster (B) is a managed service that provides a scalable and secure compute environment for running machine learning workloads. ML Compute Cluster allows you to deploy and manage a cluster of virtual machines that can be used for batch inference tasks. It supports autoscaling, so you can dynamically adjust the resources allocated to the cluster based on the demand. Additionally, it provides a range of hardware configurations to choose from, including CPU and GPU instances, which can be selected based on the requirements of the batch inference model.
HDInsight (C) is a fully-managed cloud service that makes it easy to process big data using popular open-source frameworks, such as Hadoop, Spark, and Hive. HDInsight can be used for running batch inference tasks, but it may not be the best option for deploying a single batch inference model since it is designed for processing large datasets.
Local compute (D) refers to running the batch inference model on a local machine or on-premises infrastructure. While this option may be suitable for development or testing, it may not be the best option for production use since it may not provide the scalability and reliability needed for running batch inference tasks at scale.
Based on the above analysis, the most suitable option for deploying the batch inference model for production use would be Azure ML Compute Cluster (B). It provides a scalable and secure compute environment with a range of hardware configurations to choose from, which can be adjusted dynamically based on the demand.