References for HPC on Microsoft Azure Cloud
References for High Performance Computing solutions on Microsoft Azure Cloud are provided below. This blog entry serves as URL references only. I will be providing details about these individual services in series of blog entries.
Similar to how AWS Cloud and Google Cloud are offering HPC services, Microsoft Azure Cloud offers multiple solutions to cater to different set of customers deploying their use cases on cloud. AWS Cloud references are provided here in this blog entry and Google Cloud are provided here.
As part of Microsoft Azure Cloud HPC services, they have divided the components as platform services and each of the component has different set of products inside them.
URL for reference: https://azure.microsoft.com/en-us/solutions/high-performance-computing/#platform-services
Following are provided as Platform services,
1. Network
InfiBand:
As part of parallel computing architectures, Azure cloud provides Infiband services specific to message passing interface (MPI) workloads. Their documentation clearly states about using Linux remote direct memory access (RDMA) with InfiniBand service.
https://learn.microsoft.com/en-us/azure/virtual-machines/extensions/enable-infiniband
High-performance computing on InfiniBand enabled HB-series and N-series VMs are listed here, https://learn.microsoft.com/en-us/azure/virtual-machines/overview-hb-hc
Express Route:
To create private connections between Azure datacenters and infrastructure on premises or in a colocation environment.
https://azure.microsoft.com/en-us/products/expressroute/
2. Compute
H Series VM’s – https://learn.microsoft.com/en-us/azure/virtual-machines/sizes-hpc
These instances are used specifically for Memory utilized applications.
N Series VM’s – https://learn.microsoft.com/en-us/azure/virtual-machines/sizes-gpu
These instances are GPU optimized.
3. Storage
Azure HPC Cache
https://azure.microsoft.com/en-us/products/hpc-cache/
Azure NetApp Files
https://azure.microsoft.com/en-us/products/netapp/
Cray ClusterSor
https://azure.microsoft.com/en-us/blog/supercomputing-in-the-cloud-announcing-three-new-cray-in-azure-offers/
4. Application Services
Create and manage applications and schedule jobs with Azure Batch.
https://azure.microsoft.com/en-us/products/batch
Dynamically provision Azure HPC clusters with Azure CycleCloud.
5. Intelligent Services
next generation of applications with predictive analysis using Data Lake Analytics.
https://azure.microsoft.com/en-us/products/data-lake-analytics
Gain insights that lead to better decisions by building and running machine learning models on top of your HPC data.
https://azure.microsoft.com/en-us/products/machine-learning
I have never tried Azure services specific to HPC but their H/N instances looks great in terms of handling the HPC load. I have to try them for sure after validating other cloud services.