Azure infrastructure as a service (IaaS) for every workload

This week at Microsoft Ignite, we announced several important additions to our Azure infrastructure as a service (IaaS) portfolio.

Many companies, including GEICO, H&R Block, and CONA Services, rely on Azure to run a very diverse set of business-critical workloads, often requiring dynamic and scalable infrastructure that delivers unparalleled performance.

In order to meet the needs of this diverse and growing set of mission-critical workloads that call Azure home, our infrastructure services continue to evolve to optimize the experience of running these workloads.

Comprehensive infrastructure solutions: Flexibility and choice

We announced several new offerings that expand our portfolio of available virtual machine (VM) instance sizes for general purpose, memory-intensive, and remote visualization scenarios, including the ability to run VMware environments natively and enhancements to the platform that make it even easier to migrate your workloads to Azure.

Ea v4, Eas v4, Da v4, and Das v4 series Microsoft Azure Virtual Machines now available

After being the first global cloud provider to announce the preview of Azure Virtual Machines based on the AMD EPYC™ 7452 processor, we’ve been working together with our technology partners, including AMD, to continue bringing the latest innovation to enterprises. 

This week we’re announcing the availability of the Da v4 and Das v4 Azure Virtual Machine series for general purpose Linux and Windows applications, and the Ea v4 and Eas v4 Azure Virtual Machine series for memory-intensive Linux and Windows workloads.

These new Azure Virtual Machines feature the latest AMD EPYC™ 7452 processor and up to 96 vCPUs, 672 GiBs of RAM, and 2,400 GiBs of SSD-based temporary storage. The Das-series and the Eas-series Virtual Machines support Azure Premium SSDs and will include Ultra Disk support in the near future.

New NVv4 series Azure Virtual Machines preview available

We are also enhancing our compute portfolio for Windows Virtual Desktops and high-performance computing (HPC) workloads with the preview of NVv4. These new Azure Virtual Machines feature the latest AMD EPYC™ 7742 processor and will be the first visualization-optimized Azure Virtual Machine to offer AMD RADEON INSTINCT™ MI25 GPUs. NVv4 (currently in preview) offers enhanced GPU resourcing flexibility, giving customers more choice by offering partitioned GPUs built using industry-standard SR-IOV technology. Customers can select the right size of GPU Virtual Machines with as little as 2GB of dedicated GPU frame buffer for an entry-level desktop in the cloud, and up to the whole GPU with 16GB of frame buffer to provide powerful engineering workstations. This makes entry-level and low-intensity GPU workloads more cost-effective while still giving customers the option to scale up to full-GPU processing power delivered by AMD RADEON INSTINCT™ MI25 GPUs.

Azure VMware Solutions now available in West Europe

We’re also announcing the availability of Azure VMware Solutions in the West Europe Azure region. If you are currently managing an on-premises VMware environment, Azure VMware Solutions delivers the ability to run your VMware environment natively on Azure. This gives you the option to leverage your existing VMware skills and investments while taking full advantage of the scale and automation Azure offers. Azure VMware Solutions is now supported in East US, West US, and West Europe regions.

New Azure Migrate features to streamline migration

Azure Migrate is a central hub for all your migration needs and now delivers new capabilities to accelerate the migration of physical servers and virtual machines. We have also made enhancements to the Server Assessment capabilities that reduce friction through agentless discovery options. And to ensure you have the information you need for migration; we now provide deeper application dependency analysis. Refer to the documentation for more details.

A dynamic and scalable infrastructure for uncompromised performance

One of the most valuable promises of cloud infrastructure is the ability to meet evolving business and IT requirements. In our mission to continuously improve customers’ access to dynamic and scalable infrastructure, we’ve made a couple of important additions to our portfolio.

Azure generation 2 virtual machines now generally available

Generation 2 virtual machines are now generally available on Azure. Generation 2 VMs provide support for Intel Software Guard Extensions (Intel SGX), UEFI boot architecture, and the ability to provision large VMs (up to 12TB) and OS Disks sizes that exceed 2TB.  

Generation 2 VMs are fully supported in the portal, CLI, and PowerShell interfaces, and customers can opt to use them during the provisioning and deployment process, depending on their needs. Please refer to the Windows and Linux documentation for more information.

New Azure Virtual Machine Scale Sets features now in preview

We’re also introducing the preview of new features for Azure Virtual Machine Scale Sets that will greatly simplify the experience of running virtual machines at scale, as well as improve the runtime capabilities and performance of these workloads. 

In addition to supporting a homogeneous set of VMs for a scalable app layer, you can now create an empty virtual machine scale set and add various VMs (even those belonging to different VM series) later during the VM creation process. This will allow you to achieve high availability, for example, by deploying a set of virtual machines to a single availability zone or across different fault domains in an availability zone. You can now use a Virtual Machine Scale Set to deploy a SQL high availability (HA) cluster with high availability in a zone. This will provide the high availability of SQL primary, secondary, and witness VMs in unique fault domains while maintaining the lower inter-VM network latency that is seen within an availability zone.

You can now also provision VMs with custom images using the Azure Shared Image Gallery, which provides a quick, easy and scalable way to share images across different VMs and also accelerates provisioning times.

You can also specify a scale-in policy that gives you control over the order in which VMs should be de-provisioned. Termination notifications now give customers up to 15 minutes to perform any clean-up or other pre-shutdown tasks before VMs are deprovisioned, and you can now use instance protection from scale-in to designate VMs that should not be deprovisioned during a scale-in action. 

All these new features will help you get your applications up and running quickly while giving you additional control over how your applications can scale to meet your requirements. 

HBv2 Azure Virtual Machines for HPC workloads coming soon

HBv2 VMs are designed to deliver supercomputer-class performance, message passing interface (MPI) scalability, and cost efficiency for a variety of real-world HPC workloads. HBv2 Virtual Machines support up to 80,000 cores for single MPI jobs to deliver performance that rivals some of the world’s largest and most powerful bare metal supercomputers.

Updated NDv2 Azure Virtual Machines preview

The NDv2-series Virtual Machines, currently in preview, are the latest, fastest, and most powerful addition to the GPU family, specifically designed for the cutting edge demands of distributed HPC, AI, and machine learning workloads. These VMs feature 8 NVIDIA Tesla V100 NVLINK interconnected GPUs with 32 GB of memory each, 40 non-hyperthreaded Intel Xeon Platinum 8168 processor cores, and 672 GiB of system memory. The NDv2-series Virtual Machines (currently in preview) also feature 100 Gb/sec EDR InfiniBand with support for standard Mellanox OFED drivers and all MPI types and versions. With total of 256 GB of GPU memory and 100 Gb/sec InfiniBand interconnect NDv2-series Virtual Machines are ready for the most demanding machine learning models and distributed AI training workloads utilizing CUDA, TensorFlow, Pytorch, Caffe, and other frameworks.

Proximity placement groups now generally available

A proximity placement group is a logical grouping capability for Azure Virtual Machines that you can use to decrease the network latency between a set of virtual machines. When you assign your virtual machines to a proximity placement group, their placement is optimized to deliver lower latency for your latency-sensitive workloads. We’ve seen robust customer adoption of this new feature during the preview over the last few months, and we’re pleased to now make Proximity Placement Groups generally available in most Azure regions. Please check the documentation for more information.

Azure Spot Virtual Machines

Finally, Azure Spot Virtual Machines, which give you access to unused Azure compute capacity at deep discounts, will be available soon. Spot Virtual Machines will be ideal for workloads that can be interrupted, providing scalability while reducing costs. You will be able to take advantage of Spot Virtual Machine pricing for Azure Virtual Machines or Virtual Machine Scale Sets (VMSS) to deploy opportunistic workloads of all sizes. We expect to preview this by early 2020.

In conclusion, there has never been a better time to run your workloads on, or to migrate to, Azure. We hope you enjoy Microsoft Ignite!

Additional Resources

Da series Azure Virtual Machines Linux and Windows documentation
Ea series Azure Virtual Machines Linux and Windows documentation
Azure Virtual Machine Scale Sets documentation
Azure generation 2 Virtual Machines documentation (Windows and Linux)

Azure. Invent with purpose.