After several years of Covid and not having physical in person events we were missing out on actually talking to people. Having discussions with the people at Microsoft, or people from other companies and get new views on challenges and solutions, and also to talk about real world examples and experiences.
This Ignite we were finally able to do that again. And while you might think you could have watched it all online, Yes.. The overall experience was a bit different and not as expected. But the interaction was very valuable. It’s a great chance to actually meet with the people we are “virtually” talking to all year.
During Ignite’s keynote with Satya Nadella the big message is “doing more with less” and one of the 5 biggest pillars Microsoft is investing in currently is “Be data-driven and optimize with Azure”. With that pilar the message is “Deliver the power of Azure everywhere”, with Azure Arc.
Lots of sessions were about these topics, in this blog we dive a bit deeper in the Azure Arc and Azure Stack HCI announcements
In regard to Azure Stack HCI news this is probably one of the biggest. If you have a Server license with active Software Assurance (SA) you can run Azure Stack HCI for Free, including AKS on HCI on top of that!
This is a big license advantage.
Let’s take an example. We have a 3-node Azure Stack HCI Cluster with dual 16 core processors.
That makes a total of 32 Cores per system, times 3, we have 96 cores in total. The monthly fee without SA is roughly 10 euros per core per month. With this 3-node 32 core cluster you are charged approximately 960 euro per month on your Azure bill. You get Azure support and endless new features for it, but you still need something like a Windows Server license with SA . If you already own it, you get Azure Stack HCI and AKS on HCI for free.
As you might know you can run Azure Kubernetes Service on Azure Stack HCI. So the managed Azure Kubernetes service is running in your own datacenter. Some might wonder why you would want that…. and there can be multiple reasons depending on your own circumstances and requirements. For example, you could develop apps in your internal infrastructure where you could build it from scratch and make it more secure first before deploying it on public AKS. Or the other way around having your production data only in your own datacenter and not in public cloud. Or simply because it’s just cheaper, or you don’t want the data to run outside of your company and I could provide a dozen more reasons. AKS on HCI is a really good option to run your apps on.
Running AKS in Azure cost you an x amount of money. Depending on the amount of workers, sizes of the workers and storage you add it adds up. Month after month year after year. After you purchased your Azure Stack HCI cluster AKS on HCI is totally free if you have the active SA. No matter how many workers, what size or the amount of storage it doesn’t cost you a single euro extra.
Keep in mind that the massive scaling cannot be done on your own cluster, but you can run a big amount of workers on your own cluster and have most of the AKS feature available at a fraction of the costs. Kubernetes version is a bit behind, and the available features on AKS on HCI are less but the development of that is going rapidly.
Other great news is AAD support is added which gives you authentication based on Azure Active Directory accounts to the API and cluster permissions instead of using Kubeconfig files. Kubeconfig files are not very user friendly and could cause a security issue when all these config files are scattered around the environment. Multi admin support has also been added so not only the user who setup the environment can add or change the AKS setup, but you can have multiple admins now.
Last update in regard to AKS on HCI is you can now run it on top of a cluster with Microsoft SDN (Software defined networking). PODS and containers still use calico or flannel but it’s a good step in the direction of having an Azure CNI like network environment on your AKS on HCI cluster.
Wait still more AKS news? Yes, and news that deserved its own chapter. During our journey with AKS on HCI we got involved with Microsoft to test, shape and provide feedback with the deployment of worker clusters through the Azure portal. So before this feature got in public preview you could only add or increase your worker clusters with PowerShell. As of Ignite 2022 you can now deploy and manage your worker clusters through Azure and through the available tooling like Azure CLI or ARM templates with Azure DevOps. This way Devs can use their existing code to deploy AKS everywhere. For example your datacenter, branch offices, Azure public, or all together.
A long-awaited feature is coming close, GPU Partitioning. Take your GPU’s in your host and slice it in multiple partitions that you can provide to your AVD pooled/personal hosts for users to take advantage. Give the heavy users an AVD workplace where they can work on their AutoCad drawings or do some photo or video editing at speed. Or give your container apps the GPU power to process graphic intense workloads. It is all possible on an Azure Stack HCI cluster.
Migrating from previous Hyper-V platforms to Windows Server Hyper-V platforms can be done by live migration. Unfortunately, that cannot be done when migrating to Azure Stack HCI. Mostly because of legal challenges this is not possible. Where moving VMs from VMWare and Hyper-V to Azure was possible with Azure Migrate, it was not possible to use it to migrate VMs from Hyper-V or Vmware to Azure stack HCI.. Until now. Microsoft made improvements/changes to Azure Migrate to support migration from Windows Server Hyper-V to Azure Stack HCI with Azure Migrate.
There is more news coming, but we cannot tell what that is yet, because it’s currently not publicly available. So stay tuned for more exciting news to come!