This white paper describes considerations for deploying the PI System in Microsoft's Azure cloud environment. Guidance for networking, storage and PI Server sizing is included. A minor update was made on May 17, 2018.
Thank you, this is helpful and I look forward to more resources on cloud deployments! It could tie in nicely with Azure's machine learning capabilities. I'm not an IT professional, but I understand you can also spin up SQL databases independent of a VM... Could this serve as an AF database or does that need to be on a VM?
Thank you for your feedback, Jacob. I'll pass your question to someone more technical than me.
We don't officially support installing the PI AF on an Azure SQL PaaS setup. AF uses some features that do not work on SQL Azure. I've found one PISquare post with some workarounds that would let you do it.
Osi to the cloud
But this deployment is not supported by OSI and I can't guarantee that TechSupport will be able to help you resolve any issues, or that you would be able to do an upgrade very easily. I'd advise against implementing it on a production PI system.
If you're interested in making PI data available for use with Azure Machine Learning, take a look at the PI Integrator for Microsoft Azure, which is capable of presenting PI data to Azure SQL Database, Azure SQL Data Warehouse, and Azure Data Lake Store.
I have been thinking about this document for some days, but i'm still unclear with the purpose of the document. Stating to provide best practices for using Azure IaaS for PI, but it does not really provide any real practice, just general statements. And then there is the case that this setup is not supported. So what is the point?
Of course there is a point, and Azure can be used quite well for development or test environments. I am not 100% sure the Devclub Live labs, VMs provided during OSIsoft Training, etc. use Azure, but if that is the case it does prove the value very well. I also use Azure for one of my Development environments.
What i would expect from such a document:
There is a lot of demand and a lot of value in e.g. Azure IaaS for PI, but to date the information, support and directions from OSIsoft are too much fragmented. So some more coherent approach would help.
Installing PI on Azure IaaS is supported by OSI. My earlier comment was in regards to using the Azure SQL PaaS setup which has some important differences from installing SQL on an virtual machine in the Azure cloud.
If you have some pointers to documentation stating that Azure is supported, then please. So far only virtualization is supported (HyperV and VMware) according to the system requirements of e.g. the PI DA: Request Rejected
Of course, techsupport is always helpful and will really try to help, but that is still not the same as supported. Of course, for non-production environments it runs fine, but i would not deploy a production environment on Azure without explicit backing from OSIsoft.
Retrieving data ...