We are excited to announce two additions to our H-series Azure virtual machines for High-Performance Computing (HPC). The new H-series VMs are designed to deliver leadership-class performance, MPI scalability, and cost efficiency for a variety of real-world HPC workloads.
First, we are introducing HB-series VMs optimized for applications driven by memory bandwidth, such as fluid dynamics, explicit finite element analysis, and weather modeling. HB VMs feature 60 AMD EPYC 7551 processor cores, 4 GB of RAM per CPU core, and no hyperthreading. The AMD EPYC platform provides more than 260 GB/sec of memory bandwidth, which is 33 percent faster than x86 alternatives and 2.5x faster than what most HPC customers have in their datacenters today.
Next, we are introducing HC-series VMs optimized for applications driven by dense computation, such as implicit finite element analysis, reservoir simulation, and computational chemistry. HC VMs feature 44 Intel Xeon Platinum 8168 processor cores, 8 GB of RAM per CPU core, and no hyperthreading. The Intel Xeon Platinum platform supports Intel’s rich ecosystem of software tools such as the Intel Math Kernel Library, and features an all-cores clock speed greater than 3 GHz for most workloads.
| Size | CPU cores | Memory: GB | Memory per CPU Core: GB | Local SSD: GiB | RDMA Network: | Azure Network |
| Standard_HB60rs | 60 | 240 | 4 | 700 | 100 Gbps | 40 Gbps |
| Standard_HC44rs | 44 | 352 | 8 | 700 | 100 Gbps | 40 Gbps |
‘r’ denotes support for RDMA. ‘s’ denotes support for Premium Storage.
Both of these new VMs share outstanding Message Passing Interface (MPI) scalability and I/O performance features.
Both new H-series VM sizes feature 100 Gb/sec EDR InfiniBand from our technology partners at Mellanox. This is the same HPC interconnect technology featured in the world’s most powerful supercomputer, and now Azure is bringing it to the public cloud for the first time. The InfiniBand fabrics backing the new H-series VMs are non-blocking fat trees with a low-diameter design for consistent low latencies. We are also introducing SRIOV for InfiniBand, which allows customers to use standard Mellanox/OFED drivers just like they would on a bare metal HPC cluster. In doing so, the new H-series VMs officially support all MPI types and versions, such as OpenMPI, MVAPICH2, Platform MPI, and Intel MPI, as well as all RDMA verbs. The new H-series also will support SparkRDMA for acceleration of the popular Spark analytics framework.
The new H-series VMs also feature an HPC-optimized hypervisor designed to prevent server host noise from impeding HPC application performance and scalability. The new hypervisor also exposes the topology of the underlying processor as it exists in silicon, meaning customers and applications can assign software threads to specific CPU cores to minimize data movement and maximize performance.
Finally, the new H-series VMs feature NVMe SSDs to deliver ultra-fast temporary storage for the full range of file sizes and I/O patterns. Using modern burst-buffer technologies like BeeGFS BeeOND, the new H-series VMs can deliver more than 200 GB/sec of peak injection I/O performance across a single Azure VM Scale Set. The new H-series VMs will also support Azure Premium Storage.
Both HB and HC VMs will be optimally supported by Azure CycleCloud, the simplest way to create, orchestrate and manage HPC clusters on Azure. Azure CycleCloud will also enable hybrid HPC deployments for customers wishing to pair the new H-series VMs with their existing on-premises clusters. The new H-series VMs will also be supported by Azure Batch for cloud-native batch processing.
HB-series VMs will initially be available later this year in the South Central US Azure region, while HC-series VMs will initially be available later this year in the West US 2 Azure region. Both HB and HC VMs will be rolled out across additional regions soon thereafter. Azure customers can enroll in the Public Preview for the HB-series VMs and the HC-series VMs by filling out this form.