Q&A

Top Tips for Architecting Your Azure Workloads to 99.95 Percent Availability

From how virtual machines can impact your Azure IaaS and Paas uptimes to what to never do, top cloud architect Peter De Tender offers his best tips and tricks for reaching 99.95 percent uptime by optimizing your SLA.

• By Becky Nagel
• 05/29/2018

While Azure is built with high availability in mind, it actually takes design work to get your Azure workloads to 100 percent (well, 99.95 percent) uptime.

And that's where Azure MVP, architect and freelance trainer Peter De Tender comes in. Ahead of his session, "Azure is 100% High-Available...Or Is It?" at TechMentor Redmond 2018, we got a chance to ask him for his top tips and tricks for getting your Azure Infrastructure as a Service (IaaS) and Platform as a Service (PaaS) workloads to that all-important 99.95 percent benchmark.

What would you say is the one biggest challenge for getting an Azure IaaS workload design past a 99 percent uptime?
The biggest challenge organizations are facing regarding Azure IaaS workloads is understanding they are not providing a 24/7 high availability. This is mainly a misconception, and optimizing the SLA [service-level agreement] to 99.95 or even better isn't all that hard, once you know what Azure services and features are available to get there.

Same question, but for Azure PaaS workload designs?
Providing better SLAs for your workloads is definitely possible by deploying PaaS solutions, instead of IaaS. But then again, by just deploying your workloads on Azure, doesn't give you a 100 percent high-availability guarantee.

So even in PaaS, understanding how to optimize not only your HA [high availability] but also understanding the DR [disaster recovery] capabilities are critical when talking about business-critical workload applications.

How do virtual machines (VMs) in Azure generally impact uptime?
VMs in Azure can be deployed in different ways, resulting in no/acceptable/extreme SLA. Depending on the applications running within the VMs, it might still be a challenge to get everything up and running 24/7, as Azure as a platform is not really application-aware.

Besides the VMs itself, or the applications within, there is also a "human" factor, and additional VM operations and management needed. Some of those tasks are not relevant when deploying similar workloads on PaaS. Or at least, Microsoft is taking care of that on their underlying platform infrastructure.

What would you consider your No. 1 best practice for those looking to increase their Azure workload uptimes?
Start with identifying the different SLAs put forward by Microsoft for the different Azure services. Map those with the business requirements, and make sure you keep cost as an equal important subject in the architectural design decisions than the technical factors.

What is one thing you've learned to never do when architecting either an Azure PaaS or IaaS workload?
My personal recommendation is to "skip" everything that is offering "basic" service levels. This can be within VM deployments, underlying storage, Azure App Service plans as they are -- literally -- providing basic services, which is often just not good enough for production-running workloads.

Are there any other tips you'd like to share with our audience ahead of your session at TechMentor Redmond 2018, being held at Microsoft HQ this August?
Make sure you put this session on your TechMentor agenda! I'm looking forward to having you there and answering your Azure questions!