Just getting started with this series? Make sure to read part 1: How ModelOps addresses your biggest Machine Learning challenges.
SAS and Microsoft make it easier for companies to address the challenges of machine learning model deployment, monitoring and governance. Specifically, SAS and Microsoft have built integrations between SAS® Model Manager and Microsoft Azure Machine Learning. Both are hubs for your ModelOps processes, but they also offer complementary capabilities that, when used together, make it possible to conduct ModelOps with the benefit of streamlined workflow management.
SAS Model Manager is an enterprise-ready solution for ModelOps. It provides an industry-leading framework for ModelOps that enables collection, testing, deploying, monitoring, governance, and retraining of models, regardless of what language they are written in (such as SAS, Python or R). Once you build a model, you can register it in SAS Model Manager to test, deploy and monitor performance in its production environment.
SAS Model Manager adds value to the ModelOps process by providing a common application interface that enables both business insight and IT governance, which:
• Simplifies process workflow and automation efforts.
• Improves communication and collaboration between business, IT and data scientists throughout the analytics life cycle.
• Gives business users a more active role in model monitoring and management processes.
• Governs data and assets with transparency and DevOps conformity.
• Integrates with decisioning and streaming technologies.
Azure Machine Learning is Microsoft’s enterprise-grade service for end-to-end machine learning, empowering data scientists and developers to build, deploy and manage high-quality models faster and with confidence. You can use Azure Machine Learning to accelerate time to value with industry-leading interoperability with open-source tools and frameworks. Azure Machine Learning also empowers ModelOps to bring machine learning models into production in a highly secure and tightly governed production environment. Once deployed, you can also scale models to use anywhere in the Microsoft Azure cloud, any hybrid cloud or on-premises via Azure Arc to easily meet business demands.
Additionally, Azure Machine Learning can access enterprise data and resources organizations have running in Microsoft Azure. For example, you can centrally store and manage your enterprise data in Azure Data Lake and even use Power BI to enrich your dashboards with predictive analytics, Azure Synapse to use these ML models via in-database queries, or Azure Synapse Analytics’ Spark runtime so you do not have to pay to transfer data between SAS and Microsoft environments.
Finally, Azure Machine Learning gives data scientists services to accelerate and automate their day-to-day workflows. For example, application developers gain services for integrating models into applications and platform developers have a robust set of services, backed by Azure Resource Manager APIs, for building advanced ML services. And because all of this is powered by Azure, you can use familiar management and security controls to keep your data secure, manage your costs, and more.
It is important to note that both SAS Model Manager and Azure Machine Learning are hubs for ModelOps processes. But they also complement one another in important ways:
• Azure Machine Learning adds volume and speed through scalable deployment of models to your favorite applications in the Microsoft Cloud.
• At the same time, SAS Model Manager provides an enterprise view of all models in production so that quick action can be taken to administer and replace non-performant models quickly and efficiently. As a model repository SAS Model Manager supports collaboration across data science, IT and business users with natural language explanations of outputs and a unique no-code/low-code environment that makes it easier to participate in the ModelOps process.
• Using the SAS Workflow Manager, which is natively integrated with SAS Model Manager, you can automate and delegate tasks across the model life cycle, including registering the model, running a test or validation, deploying a model, setting alerts for model accuracy declines and retraining models. You can also “push” a model from the development environment (SAS or open source) to SAS Model Manager, where all the files related to the model (score code, metadata, etc.) are automatically created and stored, ready for testing and deployment.
Next time you will learn about what ModelOps with SAS and Microsoft looks like in practice with five best practices for ModelOps with SAS Model Manager and Azure Machine Learning.
To learn more about ModelOps and our partnership with Microsoft, see our whitepaper: ModelOps with SAS Viya on Azure.