Let’s get right to the point…the Nutanix Files Storage 5.0 release includes significant improvements in performance with our latest release. In particular, we had some very notable improvements in read performance.  Why does this matter? Well, as technology evolves and demands for high-performance storage increase, continuous evaluation becomes crucial in the industry. Artificial intelligence (AI) and machine learning (ML) are driving forces behind this evaluation, both now and in the foreseeable future. The benefits from increases in performance can dramatically benefit the ability to drive new breakthroughs and insights. While today it’s AI/ML, the future can and will be something completely different and these improvements could be applied to performance requirements in general over time.

AI encompasses a diverse array of specific approaches and algorithms applied to new applications and workloads. Already, and increasingly so, AI is wielding a transformative impact on various sectors, such as automation, healthcare, natural language processing, transportation, personalized experiences, cybersecurity, and science. Historically, AI has excelled in narrow and focused use cases, notably in enhancing personalized buying recommendations, supporting real-time security analysis of business transactions to combat fraud, expediting financial application approvals, and exploring  emerging applications like Generative AI (Gen AI).

Regardless of your reasons for leveraging AI, a common thread among modern AI applications is their reliance on massive data sets. Ensuring optimal performance and value generation from these AI applications requires fast and efficient access to storage.

The typical path for accessing data involves the requests traversing through multiple layers, including CPU, memory, network drivers and other layers in the solution stack. Oftentimes, this path results in bottlenecks, typically at the CPU level. As a result, such bottlenecks can lead to delays for high-performance applications, thereby reducing efficiency. These delays can trigger cascading effects that extend to other parts of the infrastructure, such as GPUs, where ROI is gained by keeping the platform consistently busy.

As a software-defined solution, Nutanix Files Storage is well-suited to support these new performant workload requirements. With our latest release, we’re dedicated to unlocking the full potential of the connection between an application and its data. Here’s a closer look at the enhancements:

  • Support for iSCSI Extensions for RDMA (iSER), a technology that optimizes the connection between user VMs (ie: UVMs) and the CVM on the AHV hypervisor leveraging RDMA. Within the Nutanix context, the File Server VMs (FSVMs) that make up a File Server constitute the UVMs in this setup. When utilizing iSER the traffic will flow through a fast path that allows direct memory access to the target without interrupting the CPU. 

    When configured for high performance, the FSVMs will leverage hypervisor-backed virtual disks which accelerates performance over iSCSI backed virtual disks. But where does iSER fit into the picture?  This is where our software-defined platform, powered by our AHV hypervisor, offers significant advantages.

    The interaction between AHV and the CVM leverages iSER for the interactions between the FSVMs and their storage. This will reduce overhead for data processing with the CVM. The outcome is a decrease in  overall CPU usage and improvements in performance scaling.

  • Dynamically leveraging backend storage for performance is a feature already integrated into Nutanix Files Storage, facilitating capacity expansion within the file server. Initially, we start with a small number of virtual disks (7) plumbed to each file server VM, organized within volume groups that serve as the storage pool for file server data. As needed, we will expand horizontally (ie: incorporate more virtual disks per file server vm) when the backend storage reaches a certain capacity threshold, providing additional capacity. Sort of like the checkout at a grocery store, depending on how busy things are more checkout lanes are opened to ensure that customers can check out faster. This expansion is done automatically without operator intervention in this case.

    With Files Storage  5.0, when configured for high performance profiles, we deploy even  more virtual disks (11) per volume group at deployment. This distributes the workload across a larger number of storage resources, thereby increasing performance. The same dynamic expansion for capacity remains available, but now instead of just focusing on storage capacity, we offer a flexible platform  deployable to meet your diverse business needs.

  • Enhanced network capacity is pivotal in driving extensive data processing capabilities. This translates to increased business outcomes and higher ROI. The upgrades in Files Storage 5.0 will allow for leveraging the full bandwidth available on a platform containing high speed nics (ie: 100Gbps) in a node by leveraging additional networking processing queues that can be made available to the file server. This will be accomplished by increasing the number of vNIC queues available to the FSVMs, allowing for up to 8. Think of this like a highway that adds more lanes to improve traffic flow.


    Similar to the dynamic expansion above, this isn’t something an operator needs to take action to leverage. Instead, the system will recognize that a high performance configuration has been enabled and set the back-end network queues automatically.

  • Increase the support for more compute and memory for our file server VMs - Prior to this release, Files Storage  supported configuring up to 24 vCPUs in the user interface for each FSVM that made up a File Server. We will now support up to 32 vCPUs for each FSVM. Additionally, Files Storage 5.0 will now support up to 512GB of memory tied to the FSVM when using 32 vCPUs per FSVM.

Collectively, these advancements will help drive upwards of 10GB/s throughput per node for sequential reads, roughly doubling the performance from our prior release. Furthermore, as a software-defined platform, our solution offers unparalleled flexibility across a myriad of environments. This could be a modern application running in the core datacenter, handling inference tasks at the edge, or  managing clusters ranging from large petabyte-scale setups to small single-node configurations, our platform seamlessly accommodates a diverse range of requirements. Moreover, Nutanix isn’t just a special purpose storage provider focused on a singular task. We offer a full platform that can provide storage for your modern workloads like AI/ML, but can also run your modern workloads. With the ability to provide a single data and control plane, we can easily manage data from edge to core to cloud.

We are continually exploring how we can improve performance and help our customers move their businesses forward. If you are interested in these new capabilities, reach out to your Nutanix Partner or local Nutanix Account team to get more information. And with that, may the performance be with you!

Additional Resources  


© 2024 Nutanix, Inc. All rights reserved. Nutanix, the Nutanix logo,  and all Nutanix product, feature and service names mentioned herein are registered trademarks or trademarks of Nutanix, Inc. in the United States and other countries.  Other brand names mentioned herein are for identification purposes only and may be the trademarks of their respective holder(s). This post may contain links to external websites that are not part of Nutanix.com. Nutanix does not control these sites and disclaims all responsibility for the content or accuracy of any external site. Our decision to link to an external site should not be considered an endorsement of any content on such a site. Certain information contained in this post may relate to or be based on studies, publications, surveys and other data obtained from third-party sources and our own internal estimates and research. While we believe these third-party studies, publications, surveys and other data are reliable as of the date of this post, they have not independently verified, and we make no representation as to the adequacy, fairness, accuracy, or completeness of any information obtained from third-party sources.

This post may contain express and implied forward-looking statements, which are not historical facts and are instead based on our current expectations, estimates and beliefs. The accuracy of such statements involves risks and uncertainties and depends upon future events, including those that may be beyond our control, and actual results may differ materially and adversely from those anticipated or implied by such statements. Any forward-looking statements included herein speak only as of the date hereof and, except as required by law, we assume no obligation to update or otherwise revise any of such forward-looking statements to reflect subsequent events or circumstances. Any future product or product feature information is intended to outline general product directions, and is not a commitment, promise or legal obligation for Nutanix to deliver any functionality. This information should not be used when making a purchasing decision.