HomeCloud ComputingAzure Managed Lustre—Parallel file system for HPC and AI workloads | Azure...

Azure Managed Lustre—Parallel file system for HPC and AI workloads | Azure Weblog and Updates

At present, Microsoft is saying the general public preview of Azure Managed Lustre, a brand new addition to the storage choices in our Azure HPC + AI resolution. Lustre is an open-source parallel file system famend for high-performance computing (HPC) and is adept at large-scale cluster computing. Azure Managed Lustre gives the high-performance storage of Lustre with the management and consistency of Microsoft’s cloud. Consequently, clients can concentrate on their enterprise objectives, whether or not that’s constructing a fraud detection system primarily based on SAS analytics, decoding the human genome to create the following breakthrough in medication, main the frontiers of vitality exploration with seismic information processing and sustainable vitality resolution, or predicting the local weather and climate patterns impacting human life on planet earth.

Lustre as an Azure Managed Service

With Lustre as a managed service, customers can now concentrate on constructing and working their HPC and AI purposes with out managing a complicated parallel file system. HPC and AI workloads can seamlessly migrate to the cloud, retaining their compatibility with Lustre and defending present automation and platform investments developed when beforehand run on premises. Azure Managed Lustre permits a quick on-demand deployment of clusters serving international areas, assuaging advance planning whereas assembly the compliance and information residency necessities. With a palette of efficiency choices, Azure Managed Lustre delivers an elastic resolution the place customers can deploy unbiased and unique clusters with predictable efficiency eliminating the noisy-neighbor drawback generally skilled within the on-premises shared infrastructures.

The general public preview contains two sturdy SSD-based SKUs which ship a alternative in efficiency choices for mission crucial workloads: 125 MB/s and 250 MB/s for each provisioned TiB of capability. Throughout the preview, you possibly can create cluster sizes as much as 128 TiB by default, with an choice to scale as much as 768 TiB upon request. Azure Managed Lustre is constructed on the extremely sturdy Azure managed disks with locally-redundant storage consisting of three replicas so even when one or two replicas expertise points you continue to have tolerance in opposition to failures. Azure Managed Lustre delivers POSIX-compliant Lustre model 2.15 (LTS), which provides a number of efficiency enhancements. We targeted on capabilities that allow customers to devour Azure Managed Lustre simply with out worrying about how one can construct and deploy the Lustre purchasers on their compute VMs and containers. With the Azure Managed Lustre preview, clients can select from:

  • Use Azure HPC photographs prebuilt with Lustre shopper packages for Ubuntu 18.04 and 20.04 (or)
  • Obtain Lustre shopper packages from packages.microsoft.com for Linux distros – Ubuntu 18.04, 20.04 & 22.04; RHEL 7.8 & RHEL 8.0

Information tiering utilizing hierarchical storage administration permits customers to import and export information between Azure Managed Lustre and Azure Blob. This functionality permits customers to outline information archival, retention and safety primarily based on pre-defined insurance policies. Customers of enormous information units can import the information which is scorching and related for his or her lively information processing in Azure Managed Lustre clusters and archive/retain the remaining in Azure Blob. This permits them to maintain their run-time prices low. Moreover, information tiering to Azure Blob permits the customers to leverage its international presence and availability to instantiate Azure Managed Lustre clusters in a number of areas on demand. Integration with Azure Blob moreover facilitates multiprotocol information entry through NFS, HDFS and REST. Trendy containerized purposes in AI/machine studying (ML) and analytics can now run on Azure Kubernetes Service leveraging the CSI driver for Azure Managed Lustre. For instance, analytics purposes constructed on prime of SAS Viya can seamlessly leverage the excessive per-core storage efficiency that they require by integrating with the Azure Managed Lustre CSI driver.

Objective-built for HPC and AI workloads

We’re grateful to our non-public preview clients who helped us construct and excellent the product whereas addressing their key wants. We’re dedicated to constructing and supporting Azure Managed Lustre because it helps you along with your journey of working your high-performance purposes in Azure.

Bizdata logo

We work with clients who’ve many several types of HPC workloads, and we regularly hit storage efficiency constraints as a result of their tasks can contain something from hundreds to tens of hundreds of recordsdata, all of which should be learn and written in parallel. After we encounter storage bottlenecks or throttling errors on our HPC platforms, we all know it’s time to flip to Azure Managed Lustre. With Azure Managed Lustre we will add extraordinarily performant storage that may readily sustain with the HPC compute, and robotically sync enter and outcomes information with persistent blob storage. As a result of it may be began and stopped on demand, inside a couple of minutes, Azure Managed Lustre is kind of value environment friendly and pays for itself by giving us the power to scale up or scale out our HPC compute, producing outcomes a lot sooner.”—Felipe Ayora, Director—Analysis and Superior Computing, BizData.

Sharpreflections logo

The brand new Azure Managed Lustre file system is an actual sport changer. It accelerates HPC deployment of our software program and eliminates the necessity to run persistent storage servers. Setup is straightforward, and parallel i/o efficiency is quick sufficient to help very high-throughput workloads.”—Invoice Shea, CEO Sharp Reflections.

Subsequent Steps

To trial Azure Managed Lustre without cost, full the registration type. Be taught extra about how one can use Azure Managed Lustre and its varied supported options from our documentation.

Be taught extra about HPC and AI options


Most Popular

Recent Comments