Achieving high availability and meeting high SLA requirements is tough. Azure provides lots of features that enable high availability for applications, from the host and guest OS to applications using its PaaS. Architects can use these features to get high availability in their applications using configuration instead of building these features from scratch or depending on third‑party tools.
In this section, we will look at the features and capabilities provided by Azure to make applications highly available. Before we get into the architectural and configuration details, it is important to understand concepts related to Azure’s high availability.
Concepts
The fundamental concepts provided by Azure to attain high availability are as follows:
- Availability sets
- The fault domain
- The update domain
- Availability zones
As you know, it’s very important that we design solutions to be highly available. The workloads might be mission‑critical and require highly available architecture. We will take a closer look at each of the concepts of high availability in Azure now. Let’s start with availability sets.
Availability sets
High availability in Azure is primarily achieved through redundancy. Redundancy means that there is more than one resource instance of the same type that takes control in the event of a primary resource failure. However, just having more similar resources does not make them highly available. For example, there could be multiple VMs provisioned within a subscription, but simply having multiple VMs does not make
them highly available. Azure provides a resource known as an availability set, and having multiple VMs associated with it makes them highly available. A minimum of two VMs should be hosted within the availability set to make them highly available. All VMs in the availability set become highly available because they are placed on separate physical racks in the Azure datacenter. During updates, these VMs are updated one at a time, instead of all at the same time. Availability sets provide a fault domain and an update domain to achieve this, and we will discuss this more in the next section. In short, availability sets provide redundancy at the datacenter level, similar to locally redundant storage.
It is important to note that availability sets provide high availability within a datacenter. If an entire datacenter is down, then the availability of the application will be impacted. To ensure that applications are still available when a datacenter goes down, Azure has introduced a new feature known as availability zones, which we will learn about shortly.
If you recall the list of fundamental concepts, the next one in the list is the fault domain. The fault domain is often denoted by the acronym FD. In the next section, we will discuss what the FD is and how it is relevant while designing highly available solutions.
The fault domain
Fault domains (FDs) represent a group of VMs that share a common power source and network switch. When a VM is provisioned and assigned to an availability
set, it is hosted within an FD. Each availability set has either two or three FDs by default, depending on the Azure region. Some regions provide two, while others provide three FDs in an availability set. FDs are non-configurable by users.
When multiple VMs are created, they are placed on separate FDs. If the number of VMs is more than the FDs, the additional VMs are placed on existing FDs. For example, if there are five VMs, there will be FDs hosted on more than one VM.
FDs are related to physical racks in the Azure datacenter. FDs provide high availability in the case of unplanned downtime due to hardware, power, and network failure. Since each VM is placed on a different rack with different hardware, a different power supply, and a different network, other VMs continue running if a rack snaps off.
The next one in the list is the update domain.
The update domain
An FD takes care of unplanned downtime, while an update domain handles downtime from planned maintenance. Each VM is also assigned an update domain and all the VMs within that update domain will reboot together. There can be as many as 20 update domains in a single availability set. Update domains are non-configurable by users. When multiple VMs are created, they are placed on separate update domains. If more than 20 VMs are provisioned on an availability set, they are placed in a round‑robin fashion on these update domains. Update domains take care of planned maintenance. From Service Health in the Azure portal, you can check the planned maintenance details and set alerts.
In the next section, we will be covering availability zones.
Availability zones
This is a relatively new concept introduced by Azure and is very similar to zone redundancy for storage accounts. Availability zones provide high availability within a region by placing VM instances on separate datacenters within the region. Availability zones are applicable to many resources in Azure, including VMs, managed disks, VM scale sets, and load balancers. The complete list of resources that are supported by availability zones can be found at https://docs.microsoft.com/azure/availability‑ zones/az‑overview#services‑that‑support‑availability‑zones. Being unable to configure availability across zones was a gap in Azure for a long time, and it was eventually fixed with the introduction of availability zones.
Each Azure region comprises multiple datacenters equipped with independent power, cooling, and networking. Some regions have more datacenters, while others have less. These datacenters within the region are known as zones. To ensure resiliency, there’s a minimum of three separate zones in all enabled regions. Deploying VMs in an availability zone ensures that these VMs are in different datacenters and are on different racks and networks. These datacenters in a region relate to high‑speed networks and there is no lag in communication between these VMs. Figure 2.1 shows how availability zones are set up in a region:
You can find more information about availability zones at https://docs.microsoft.com/ azure/availability‑zones/az‑overview.
Zone‑redundant services replicate your applications and data across availability zones to protect from single points of failure.
If an application needs higher availability and you want to ensure that it is available even if an entire Azure region is down, the next rung of the ladder for availability is the Traffic Manager feature, which will be discussed later in this chapter. Let’s now move on to understanding Azure’s take on load balancing for VMs.
Load balancing
Load balancing, as the name suggests, refers to the process of balancing a load among VMs and applications. With one VM, there is no need for a load balancer because the entire load is on a single VM and there is no other VM to share the load. However, with multiple VMs containing the same application and service, it is possible to distribute the load among them through load balancing. Azure provides a few resources to enable load balancing:
- Load balancers: The Azure load balancer helps to design solutions with high availability. Within the Transmission Control Protocol (TCP) stack, it is a layer 4 transport‑level load balancer. This is a layer 4 load balancer that distributes incoming traffic among healthy instances of services that are defined in a load-
balanced set. Level 4 load balancers work at the transport level and have network‑ level information, such as an IP address and port, to decide the target for the incoming request. Load balancers are discussed in more detail later in this chapter.
- Application gateways: An Azure Application Gateway delivers high availability to your applications. They are layer 7 load balancers that distribute the incoming traffic among healthy instances of services. Level 7 load balancers can work at the application level and have application‑level information, such as cookies, HTTP, HTTPS, and sessions for the incoming request. Application gateways are discussed in more detail later in this chapter. Application gateways are also used when deploying Azure Kubernetes Service, specifically for scenarios in which ingress traffic from the internet should be routed to the Kubernetes services in the cluster.
- Azure Front Door: Azure Front Door is very similar to application gateways; however, it does not work at the region or datacenter level. Instead, it helps in routing requests across regions globally. It has the same feature set as that
provided by application gateways, but at the global level. It also provides a web application firewall for the filtering of requests and provides other security- related protection. It provides session affinity, TLS termination, and URL-based routing as some of its features.
- Traffic Manager: Traffic Manager helps in the routing of requests at the global level across multiple regions based on the health and availability of regional endpoints. It supports doing so using DNS redirect entries. It is highly resilient and has no service impact during region failures as well.
Since we’ve explored the methods and services that can be used to achieve load balancing, we’ll go ahead and discuss how to make VMs highly available.
VM high availability
VMs provide compute capabilities. They provide processing power and hosting for applications and services. If an application is deployed on a single VM and that machine is down, then the application will not be available. If the application is composed
of multiple tiers and each tier is deployed in its own single instance of a VM, even downtime for a single instance of VM can render the entire application unavailable. Azure tries to make even single VM instances highly available for 99.9% of the time, particularly if these single‑instance VMs use premium storage for their disks. Azure provides a higher SLA for those VMs that are grouped together in an availability set. It provides a 99.95% SLA for VMs that are part of an availability set with two or more VMs. The SLA is 99.99% if VMs are placed in availability zones. In the next section, we will be discussing high availability for compute resources.
Compute high availability
Applications demanding high availability should be deployed on multiple VMs in the same availability set. If applications are composed of multiple tiers, then each tier should have a group of VMs on their dedicated availability set. In short, if there are three tiers of an application, there should be three availability sets and a minimum of six VMs (two in each availability set) to make the entire application highly available.
So, how does Azure provide an SLA and high availability to VMs in an availability set with multiple VMs in each availability set? This is the question that might come to mind for you.
Here, the use of concepts that we considered before comes into play—that is, the fault and update domains. When Azure sees multiple VMs in an availability set, it places those VMs on a separate FD. In other words, these VMs are placed on separate physical racks instead of the same rack. This ensures that at least one VM continues to be available even if there is a power, hardware, or rack failure. There are two or three FDs in an availability set and, depending on the number of VMs in an availability set, the VMs are placed in separate FDs or repeated in a round‑robin fashion. This ensures that high availability is not impacted because of the failure of the rack.
Azure also places these VMs on a separate update domain. In other words, Azure tags these VMs internally in such a way that these VMs are patched and updated one after another, such that any reboot in an update domain does not affect the availability of the application. This ensures that high availability is not impacted because of the VM and host maintenance. It is important to note that Azure is not responsible for OS‑level and application maintenance.
With the placement of VMs in separate fault and update domains, Azure ensures that all VMs are never down at the same time and that they are alive and available for serving requests, even though they might be undergoing maintenance or facing physical downtime challenges:
So far, we have discussed achieving high availability for compute resources. In the next section, you will learn how high availability can be implemented for PaaS.