High availability for VMs on azure


Increased availability of applications / services is one of the most important reasons why organizations move to cloud. So what is availability ?

Per Wikipedia, availability is defined as "The ratio of (a) the total time a functional unit is capable of being used during a given interval to (b) the length of the interval".  A simpler definition would be, "availability is the proportion of time a system is in a functioning condition".

High availability obviously is good for business as it means little or no down time for applications that the business relies on. Let us see what are some of the availability options on azure for VMs.

Azure has something called regions, they are nothing but a data center or a group of data centers in a geographic region. For example Central US is a region in Azure, as is East US and West US. Each azure region can be either one or more data centers. Each region is typically paired with another region so that data, applications, vms etc can be replicated across these regions to increase overall availability.

In each region, there can be multiple availability zones. An availability zone is a data center. Most azure regions still have only one zone, but some regions like Central US have 3 zones available. What does this mean? It means in Central US Azure region, 3 separate data centers are available relatively close to each other in such a way that, they are close enough to make replication fast, but separated by enough distance so that at least one may survive any local / regional disasters / outages.
Here is a list of some azure regions and the availability zones in those regions.



So availability zone is one of the ways how azure provides high availability. In addition, there are also availability sets. If VMs are deployed into an availability set, azure ensures that they are distributed across multiple hardware servers and on different server racks in the data center. This is to ensure that in case a power or a network outage occurs, at least one other VM is still available.
Azure provides us with parameters called Fault Domains and Update Domains, which are configurable and need to be chosen correctly while deploying VMs into an availability set.

Fault Domain means a virtual grouping of hardware including server, power outlet and network. When deploying VM into an availability set, fault domains can be 1 to 3. Azure creates VMs according to the number chosen which automatically sets up the deployment to be resilient against hardware or network failures localized to any one area of the data center.

Update domain means a virtual group of hardware and software which gets updated at any one time. Number of update domains that are available for VMs range from 1 to 20. So what this means is that azure will deploy VMs into multiple update domains as well along with fault domains. When updates are rolled out, they are rolled out to one update domain at a time and even if an update causes disruption, it affects only one of the VMs and other VMs in the other update domains are still available.

So this combination of multiple availability zones, and availability sets including fault domains and update domains increases the overall availability of the VMs.

Here is a link to Microsoft's SLA for VMs on Azure. A single VM offers three nines (99.9%) availability, whereas deploying in availability sets takes it up to three nine five (99.95%) and finally the highest availability of four nines (99.99%) is when deployment is across 2 or more availability zones.

Now, what does this mean in terms of actual downtime in hours / minutes ? Let us check the chart from wikipedia

As you can see there is a significant reduction in downtime between just 99.9% and 99.99% availability.

So, using availability zones and availability sets (with appropriate number of fault domains and update domains), VMs in azure can be deployed in a highly available configuration.

References:
https://docs.microsoft.com/en-us/azure/virtual-machines/windows/regions-and-availability
https://social.technet.microsoft.com/wiki/contents/articles/51828.azure-vms-availability-sets-and-availability-zones.aspx
https://gallery.technet.microsoft.com/Set-Azure-Resource-Manager-f7509ec4


Comments

Post a Comment

Popular posts from this blog

How to enable elasticsearch access from remote machines

Image
Elasticsearch is one of the most popular open source search platforms available today. In this post, we will look at how to enable remote access to elastic. Now, what do I mean by remote access? By remote access, I mean ability to access the elastic api endpoints from machines other than where it is setup. The default elasticsearch installation is localhost only, you can access it only from the machine where it is setup. While this is okay for POC type development work, for most other use cases we need elasticsearch to be installed on a server somewhere and be able to access it from other machines. Before we get started with the steps needed to configure elasticsearch installation for remote access, do note that these are applicable only for elastic 7.1.x (maybe even 7.x). Steps: Navigate to the folder where elasticsearch is installed. It should look something like this screenshot Open up the config folder. There should be a elasticsearch.yml file. Open it. In the Netwo...
Read more

Your project does not reference error!

The error " Your project does not reference ".NETFramework,Version=v4.6.1" framework. Add a reference to ".NETFramework,Version=v4.6.1" in the "TargetFrameworks" property of your project file and then re-run NuGet restore " troubled me a bit at work today. The target framework was set correctly and I had 4.6.1 on my dev machine. IDE was VS 2019. So what was the problem?  Good old fashioned googling helped me to resolve the issue and so adding this note here as a reference for myself. It turns out the solution was to delete the obj folders and just recompile. Just doing a clean and rebuild did not help. I had to delete the obj folder manually and then recompile worked. The link that helped me with this is in the reference below.  It seems like lot of people have come across this same problem. Latest comment on that thread is from Jan 2020 and looks like the issue is still unresolved from VS team. This is at least one of the workarounds t...
Read more

Add co-administrators for your azure subscription

Image
I use my Visual Studio Enterprise subscription at work which gives me a free monthly credit of close to 10,00 INR (around 150 USD). I use it quite extensively to learn about azure services, create proof of concepts and generally play around with azure services. There was a situation recently where I had to let another user manage azure services which I had created and also allow him to use more of azure services, all using my subscription. I was wondering if there was a way to do this and if this would even be allowed, considering that I was only using the free credit obtained from VS enterprise and not really paying anything extra to use Azure. The good news is that there is! Good folks at Microsoft have provided a way to add on additional administrative users for a single subscription. Now that I think about it, for large enterprises, this would be a major requirement. No way that everything that a decent sized company has on azure would be managed by just one admin account. B...
Read more