With such a range of options, it can be tricky to know which Azure VM sizes and types are the best fit for your…
In an on-premises environment, you define the number of virtual CPUs (vCPUs) and the total amount of memory that your VM will have during the deployment process. With cloud computing, in general, you need to select a specific VM size with a fixed number of vCPUs and memory. If you plan to migrate on-premises servers to Azure, choose the VM size that most closely matches your server’s requirements for CPU, memory and storage.
There are also groups of machine sizes that are designed to accommodate a specific application. For example, some Azure VM sizes are optimized for high CPU, memory or storage usage. After you’ve made your choice, you can measure performance and adjust the size as needed.
Azure VM types
Instead of focusing on the entire list of Azure VM sizes, it’s helpful to focus on the different types available.
General purpose: This VM type is designed for a balanced ratio of CPU to memory and includes the Dsv3, Dv3, DSv2, Dv2, DS, D, Av2, A0-7 and B — which is currently in preview — VM sizes. These are good options for workloads that need to support low to medium activity levels. The basic tier within this family is best for development and test scenarios that don’t require advanced functionality, like load balancing and auto scaling.
Compute optimized: This type has a high CPU-to-memory ratio and is optimized for CPU-intensive workloads. The Azure VM sizes within this group include the Fs and F series.
Memory optimized: This VM type provides a high memory-to-CPU ratio and is a good fit for production workloads, such as database servers and in-memory caches, which require a lot of memory. Memory-optimized Azure VM sizes include the Esv3, Ev3, M, GS, G, DSv2, DS, Dv2, and D series VMs. The M-Series VM size offers up to 128 vCPUs and 2 TB of total memory.
Storage optimized: For workloads that require heavy read/write operations and low latency performance, such as big data, the Ls series is a good fit. There are four Azure VM sizes in the Ls series, ranging from hundreds to thousands of gibibytes (GiB) of local solid-state drive (SSD)-backed storage, offering fast read/write performance.
GPU: The NV and NC VM sizes in this group are for heavy graphic-rendering workloads and video editing. These Azure VM sizes are powered by Nvidia GPUs.
High performance compute: These compute-intensive VMs include the H and A8-11 series. The underlying hardware that powers these VMs are optimized for compute- and network-intensive workloads, including high-performance computing cluster applications.
Beyond CPU and memory
CPU and memory are important considerations for VMs, but there are other capabilities, such as storage, network configuration and regional support, to keep in mind.
Every VM has a virtual OS disk located in Azure Storage. By default, this disk is 127 GB in size, but you can increase it to 2,048 GB. Enterprises can also add more data disks, which can support a capacity of up to 4,095 GB each. However, the number of data disks supported varies depending on the Azure VM size. For example, a Standard_A0 VM only supports one data disk, while a Standard_GS5 supports up to 64.
Enterprises need to choose between Azure Standard and Premium Storage. The key difference between the two is that Premium Storage is backed by SSD-based disks, delivering better storage performance. If you want to deploy an I/O-intensive workload in Azure, choose Premium Storage. Any Azure VM size that includes an “s” in its name supports Premium Storage, including the DS, DSv2, GS, Ls and Fs series VMs. Larger Azure VM sizes support more I/O operations per second than their smaller counterparts.
Each Azure VM type also supports a different number of maximum network interfaces. This is an important distinction for VMs that need to have a network interface in multiple virtual networks, such as VMs that act as a network appliance.
Lastly, pay attention to where Azure VM types are available, as well as their price. Certain VM types are only available in a select number of regions. It’s also possible for the price of a VM size to be different across regions. When evaluating a VM size, make sure it is available in your chosen region, and use the Azure Pricing Calculator to estimate cost.