As mentioned in my previous blog about SAP TechEd && d-code Berlin we had several SAP Mini Code Jam sessions. However, when preparing our sessions we had several Microsoft specific requirements (obviously) that would not fit in the overall concept of other Mini Code Jam sessions.
So we decided to go with Microsoft Azure and create our own images. This actually turned out to be quite helpful, because it allowed us to literally make “last-night-changes” to the system. In the end we hosted 19 clients on Microsoft Azure which were successfully used by the participants of the three SAP Mini Code Jam sessions.
Let me quickly outline the steps that we did to get this working (because it was the first time that I was using Microsoft Azure PowerShell — I ran in some issues, but in the end it worked perfectly well)
On Microsoft Azure we used a standard Windows Server Gallery template to create our master image:
After that it was fairly easy to copy the images. Following the instructions from http://michaelwasham.com/windows-azure-powershell-reference-guide/windows-azure-virtual-machines-with-powershell-basics/ I came up with this Azure PowerShell script which allowed me to quickly create the required 19 virtual machines:
Add-AzureAccount Select-AzureSubscription -SubscriptionName "Windows Azure MSDN - Visual Studio Premium" Set-AzureSubscription -SubscriptionName "Windows Azure MSDN - Visual Studio Premium" -CurrentStorageAccount "saptechedstorage" $name = "sapteched-emea-00" $service = $name + "-srv" $image = "sapteched-emea-teched-sysprep-20141110" $pwd = "Welcome1" $un = "student" New-AzureVMConfig -Name $name -InstanceSize "Standard_D1" -Image $image | Add-AzureProvisioningConfig -Windows -AdminUserName $un -Password $pwd | New-AzureVM -ServiceName $service -Location "West Europe"
As you can see we created the images with a Standard D1 = 1 Core, 3,5 GB VM size (which we did actually scale up in some cases) and deployed them to “West Europe” (of course for SAP TechEd && d-code Las Vegas we had deployed them to “West US”). This resulted in not one single participant complaining about performance
In the end we had our 19 instances
and looking at the monitor we can clearly see when the users were working