6.0 KiB
Avere vFXT
The Avere vFXT is an enterprise-scale clustered file system built for the cloud. It provides scalability, flexibility, and easy access to data stored in the cloud, in a datacenter, or both. High-performance computing workloads are supported with automatic hot data caching close to Azure Compute resources. To learn more please visit the Avere vFXT documentation page.
Tutorials
These tutorials help you understand cluster performance testing and common use-case tasks.
- VFX and Animation Rendering with HPC Cache and Avere vFXT on Azure - Use HPC Cache or Avere vFXT as part of your burst rendering architecture. This document describes the cloud burst rendering architecture and how to build out the infrastructure using Terraform examples, modules, and a provider.
- Virtual Machine Client Implementations that mount the Avere vFXT Edge Filer - This tutorial discusses how to deploy and mount 3 types of virtual machines: loose VMs, VM availability sets (VMAS), and VM scale sets (VMSS).
- Measure HPC Cache or vFXT performance with vdbench - Deploys vdbench on an N-node cluster to demonstrate the storage performance characteristics of the HPC Cache or Avere vFXT cluster
- Data Ingestor - This tutorial implements a data ingestor containing the tools required to efficiently load data onto the Avere vFXT Edge Filer.
- Rendering using Azure Batch and HPC Cache or Avere vFXT - Demonstrates how to use the Autodesk Maya Renderer with Azure Batch and the HPC Cache or the Avere vFXT cluster to generate a rendered movie.
- Why use the HPC Cache or Avere vFXT for Rendering? - Shows the results of rending against NFS at various latencies and how HPC Cache or the Avere vFXT hides the latency.
- Best Practices for Improving Azure Virtual Machine (VM) Boot Time - The Avere vFXT is commonly used with burstable compute workloads. We hear from our customers that it is very challenging to boot thousands of Azure virtual machines quickly. This article describes best practices for booting thousands of VMs in the fastest possible time.
- Windows 10 workstation for Avere vFXT - Creates a Windows workstation within the same VNET as the Avere vFXT and automatically mounts the vFXT cluster and installs various Azure tools for debugging.
- Transfer Custom VM Image from GCE to Azure - A guide on directly transferring your custom VM image from GCE to Azure.
Resources
- vFXT guides - Additional documentation about Avere vFXT clusters
- vfxt.py usage - Usage guide for the vfxt.py script
- Azure FXT Edge Filer documentation - Information about the Azure FXT Edge Filer hybrid storage cache (released July 2019)
- FXT Cluster Creation Guide - Although this guide is for creating clusters of physical FXT appliances, some configuration information is relevant for vFXT clusters as well.
- Cluster Configuration Guide - A conceptual guide and complete settings reference for administering an Avere cluster.
- Dashboard Guide - How to use the cluster monitoring features of the Avere Control Panel.
Legal Notices
Microsoft and any contributors grant you a license to the Microsoft documentation and other content in this repository under the Creative Commons Attribution 4.0 International Public License, see the LICENSE file, and grant you a license to any code in the repository under the MIT License, see the LICENSE-CODE file.
Microsoft, Windows, Microsoft Azure and/or other Microsoft products and services referenced in the documentation may be either trademarks or registered trademarks of Microsoft in the United States and/or other countries. The licenses for this project do not grant you rights to use any Microsoft names, logos, or trademarks. Microsoft's general trademark guidelines can be found at http://go.microsoft.com/fwlink/?LinkID=254653.
Privacy information can be found at https://privacy.microsoft.com/en-us/privacystatement
Microsoft and any contributors reserve all others rights, whether under their respective copyrights, patents, or trademarks, whether by implication, estoppel or otherwise.
Contributing
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.microsoft.com.
When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.