Introducing New Performance Tiers for Azure Managed Lustre: Enhancing HPC Workloads

Introduction

In the realm of high-performance computing (HPC) and AI workloads, the need for agile and powerful storage solutions cannot be overstated. Azure Managed Lustre (AMLFS) has emerged as a game-changing solution, providing managed, pay-as-you-go file systems optimized for these data-intensive tasks. Building upon the success of its General Availability (GA) launch last month and in direct response to customer feedback that we received during our Preview period, we’re excited to unveil two new performance tiers for AMLFS, designed to cater to the diverse array of customer needs. This blog post explores the specifics of these new tiers and how they embody a customer-centric approach to innovation.

40 MB/s per TiB Option: Optimizing Cost and Capacity

The 40MB/s per TiB performance tier represents a significant milestone for Azure Managed Lustre users. It directly addresses the needs of customers dealing with larger datasets, providing a lower-cost option without compromising on performance. Built on the exceptional speed, reliability, and low-latency characteristics of Azure Managed SSDs, the 40MB/s per TiB configuration, ensures that organizations can unlock the power of HPC and AI without incurring exorbitant costs. This tier offers a default maximum file system capacity of 768 TiB. With this option, Azure Managed Lustre becomes an even more accessible solution for enterprises seeking scalable and cost-effective storage solutions.

500 MB/s per TiB Option: Tailored Performance for Massive Throughput Requirements

Azure recognizes that not all workloads require massive storage capacities. The introduction of the 500MB/s per TiB performance tier specifically speaks to the needs of customers dealing with smaller datasets. Constructed on top of Azure Managed SSDs, this tier strikes a balance between performance and capacity, ensuring that users can access a storage solution that aligns precisely with their requirements. With a minimum file system size of 4 TiB, this option empowers organizations to avoid over-provisioning on capacity to meet their performance requirements, allowing customers to efficiently manage their resources. This granular approach to performance tiers demonstrates Azure's commitment to catering to a wide spectrum of customer needs.

Pricing

The introduction of these performance tiers underscores Azure's dedication to listening to its customers and iterating its offerings to match real-world demands. While pricing details for these new options are set to be published next month, we want to be completely transparent. Below is a glimpse of the anticipated pricing across different regions for the 40 MB/s per TiB and 500 MB/s per TiB performance tiers:

[Note: Actual pricing details will be found on the official Azure pricing page next month (https://azure.microsoft.com/en-us/pricing/details/managed-lustre/)]

The introduction of the 40 MB/s per TiB and 500 MB/s per TiB performance tiers for Azure Managed Lustre marks a significant step forward in the realm of HPC and AI storage solutions. These options cater to a diverse range of workloads, from large-scale datasets to smaller, performance-intensive tasks. Azure's responsiveness to customer feedback and its focus on aligning its services with actual user needs further solidify its position as a leading cloud service provider. As the industry continues to evolve, Azure Managed Lustre remains at the forefront of delivering innovative solutions that empower organizations to thrive in the digital age.

Learn more about using Azure Managed Lustre for your HPC and AI solutions

• Learn more about Azure Managed Lustre
• Re-think your HPC project spaces by Deploying Per-User or Per-Project Lustre File Systems
• Find out how we've been Benchmarking Azure Managed Lustre
• Need to access Lustre from a Windows Client? Learn how to Access Azure Managed Lustre through a Samba server
• AMLFS + CycleCloud is a match made in HPC Heaven - Integrate CycleCloud and Azure Managed Lustre
• Visit our Azure HPC hub for more technical content developed for HPC

#AzureHPC #AzureHPCAI