A year and a half ago Microsoft began adding machine learning capabilities to its Azure cloud platform. The new platform was called Azure Machine Learning and was intended to bring together capabilities of robust algorithms and analytics tools that spanned Microsoft services. Since then, Microsoft’s worked and tweaked the service and in February of this year, offered Azure ML as a fully managed, fully supported cloud service.
Unfortunately, Microsoft’s Azure ML only rolled out to data centers in North America. With Azure ML relegated to North America, it posed a challenge for Azure customers who happened to be under regulatory constraints that require all their data storage and computation resources be located in areas outside of the U.S.
Today, Microsoft announced that it is taking steps to open Azure ML to Azure data centers, starting with deployment in Southeast Asia region (Singapore). Along with today’s announcement, Principal Program Manager at Microsoft, Hai Ning gives a few instructions to Azure ML customers in Southeast Asia to deploy Azure ML solutions in their data centers today.
Microsoft is also aware there a some limitations to its announcement, and a promising that it is actively working them.
- With this release, you can only copy experiments between workspaces that belong to the same region. In the future, we will enable copying experiments between workspaces across multiple regions.
- You can only list workspaces under one region at a time in the workspace selector. In the future, you will be able to see a full list of workspaces you have access to across all regions at the same time.
- If you use a Free workspace or a Guest Access workspace from the Studio homepage, these workspace will continue to be created and operated out of the US South Central region. In the future, you will be able to create Free and Guest Access workspaces in the region that is determined to be more optimal.
- If you deploy a web service from a predictive experiment, the web service endpoints can only live in the same region that the experiment is created in. In the future, you will have the flexibility of creating experiments in one region, and deploying generated web service endpoints into different regions.
According to Hai, Microsoft will be aggressively expanding the markets that Azure ML is available in, starting with its next target, Western Europe.Further reading: Asia, Azure, Cloud, learning, machine, Microsoft, ML