NVIDIA Corporation (NVDA) Earnings Call Transcript & Summary

Before we begin, we wanted to cover a few housekeeping items. At the bottom of your screen are multiple application widgets you can use. All the windows on your screen are resizable and movable. If you have any questions during the webcast, you can submit them through the Q&A window to the right of the slides, and we'll try to answer these at the end of the event. You can find additional answers to some common technical issues located at the help widget at the bottom of your screen. Now without further ado, we'll turn the event over to our speaker to begin the presentation.

Good morning, good evening, and hello, everyone. My name is Atif Qadeer, and I am the lead solution engineer, focusing on network and security at VMware. So today, our topic is vSphere and DPU BlueField. And I am privileged that I'm accompanied by NVIDIA and Dell Technologies in this session. So a quick disclaimer on the features available today versus what will be available tomorrow, so I'll quickly whiz through. So the session will flow in this order. So I will talk about the VMware's approach to DSEs and DPUs and all the ecosystem around it. I'll then hand it over to NVIDIA, who will talk about the joint collaboration and all the good work we are doing together as a joint venture. And then NVIDIA would hand it over to Dell Technologies who are our OEM partner and have been pioneered in the Project Monterey from inception leading to where we today. So I'll take you through the overview of Distributed Services Engine, the Data Processing Unit and all the details around it. We will then touch on the DPU and the SmartNIC ecosystem in detail, how it functions and what are the components that make up us DPU. And then we would look into the DPU-based acceleration for NSX, which is actually the primary use case. So what is in it for me? So we want you to understand what DSE is, that's Distribution Service Engine. How does the offloading networking services work? And why are we doing it? And then how can you get started? So Distributed Service Engine. As some of you may have known it as Project Monterey is now available from vSphere 8 and later, right? So Project Monterey is now DSE or Distributed Service Engine. It's a -- feature of vSphere. And the initiative aim is to deliver a new level of system architecture that can efficiently process and manage data across wide range of computing environment. So the idea is that we can simplify infrastructure and workload management, strengthen infrastructure security with an enhanced Zero-trust model where we can actually segregate the security functions from the OS and move it into the Data Processing Units, DPUs, and it boosts infrastructure performance as well. So we are touching new markets and territories in terms of where we have not been able to achieve those level of optimization throughput latency and things like that. So let's dive a little bit deeper. So today, this is how a vSphere looks like. So you have the -- in your vSphere, you have the ESXis or ESXi. And you have different services that includes NSX services, the VSAN data services, the host management and the compute hypervisor and you have a NIC, which can be classed as a performance NIC. So what we then do is we replace that NIC and introduce a DPU, which is the Data Processing Unit. And as part of the first phase, what we're doing is we are moving and offloading the NSX service into a DPU. And what really happens is there is another copy of hypervisor running inside this data processing unit. We will look at it into more detail as well. And as part of the Phase I, the vSphere 8 is already released and you can use NSX services that are offloaded to these DPUs. So later phases may include DPUs on bare metal as well. So what does it look like, right? So well, some applications are not chosen to be virtualized. They are either too old or just require a lot of resources for a single application or a VM. So what customers still want to do is take advantage of VMware products and capabilities. And this is where the DPU comes in, right? So the bare metal service running Linux or Windows will be able to have DPUs installed, running ESXi on DPU even if the x86 bare metal server is not running the VMware hypervisor. So it's a big thing for us, really. Again, remember the disclaimer slide, that's all futures. So let's look at the [indiscernible] offer data processing unit a.k.a. the DPU. So what is a DPU, right? So it looks like in NIC and it has a couple of ports that could be 25, 40 or 100 gig. And we have a programmable accelerator and a high-speed connector between the accelerator and the CPU. This is the ARM CPU. And if you recall, the 2019 VMworld sessions, we were talking about ESXi. So that was the early days of research and development where we were trialing out different permutation and combination. So what you see today from a DPU perspective comes from a long cycle of innovation and collaborative effort with NVIDIA and Dell. So what we're doing is we are installing a copy of ESXi on the flash. And you may think that if you're installing ESXi on the flash that will create a lot of logging functions and logs could actually burn out the flash as a result of read and write operations. So what we are doing is we are moving the read and write operations, which are the logs into the hypervisor, which is the x86 and that provides us improved capability from log analysis perspective because that then becomes a single point of storage or a single point of management for [ Day 2 ] functions, should we need to explore the logs in more detail. We then have the services that run on the DPU, which is this CPU. And then what we can then do is, we can expose virtual devices to the x86 server for different type of functions. So there's one catch. And you cannot buy a DPU online or install it on any x86 server. The solution requires a distributor service engine ready node. So if you remember the early days of VSAN where you also needed a VSAN Ready Node, the same principle to this. And in order to simplify operations, maintenance and communication between the x86 and the DPUs because they are 2 distinct hardware basis, 2 distinct copies of ESXi but still managed from the same vSphere. The solution introduces the need of a special card, and we call it the Daughter Card. This card has several ports, 1 port connects to BMC, which is the baseboard management controller, out-of-band or iDRAC or iLO as some people have known it and these port connect to the DPU. And in future, there may be an additional port available for dual DPUs as well. So we are not using the PCI Express to control the BMC functions for this DPU card. Like I mentioned that we have Dell and NVIDIA on this session as well. And by building a flexible and open ecosystem of DPU solution, Dell and NVIDIA are ensuring that customers have access to latest technologies and capabilities that are essential for managing data-intensive workloads. So this collaborative effort between Dell, NVIDIA and VMware enable organizations to accelerate their digital transformation journey, unlock new business opportunities and achieve better business outcomes. So a little bit around vSphere 8. So let's look at what vSphere offers today, right? So it's all about the CPU, GPUs and then DPU services. So Project Monterey was a joint collaboration between VMware, NVIDIA and Dell as an OEM partner to develop a new software-defined architecture for data centers. So the aim of the project was to provide a scalable, flexible architecture, and we have achieved that vision now as a result of vSphere 8 and the DSE services. And the idea is that we should be able to offload the core services or some of the services that -- into the hypervisor and a release that CPU utilization from that perspective. It gives us the performance benefits, agility, segregation of security functions and the management clarity in terms of single point of enforcement, single point of management. So that comes with some real-world outcomes. So there's a white paper jointly written with collaboration with NVIDIA. And here the use case was to expedite and see the performance benefits of Redis. So if you don't know, Redis is an open source powerful in-memory data structure store that provides higher performance access to data and supports wide range of data structures and features. And it's very easy to integrate with different programming languages and is widely used in modern applications and real-time systems and sometimes another data intensive use cases. So the key word here is data intensive, which means that there is a cost on the CPU on the performance. And you can see the results here. So NVIDIA is going to talk about it in detail, so I'm not going to spend more time on this. So DPU-based acceleration for NSX. So NSX if you don't know is our flagship product that basically connects all the dots together and underpins our vision for virtual cloud network. So it's basically a feature function stack that provides end-to-end connectivity into the fabric, all delivered in software because VMware is a software company. And provides security functions like distributed firewall, IDS/IPS or ATP, which is the advanced threat protection package. So if we look at the network and security functions, what we have is the ESXi, and we have NSX functions running on that ESXi. And what we do is we offload those NSX functions into the SmartNIC or the DPU, which makes the ESXi slimmer and introduces security isolation and provides [indiscernible] capabilities from that perspective. So what we have is 2 methods of communication inside an ESXi, so we have the enhanced data path and a standard data path. So enhanced data path part uses the poll mode driver. And is able to achieve a high packet per second and low latency, but it's enough to saturate 40 gig link. Similarly, a standard data path uses an interrupt mode driver and can achieve decent packet per second and latency figures but it's enough to saturate a 10 gig link, right? So those are the key demarcations in terms of both functions work. But then at the same time, we have 2 different types of capabilities as well available. So one is SR-IOV which I see a lot being used in the telco industry because of higher throughput and low latency requirements. But clearly, there are some drawbacks of SR-IOV compared to enhanced data path. And therefore, NSX has the benefit if we introduce the SmartNIC or the DPUs. So in this slide, what we are trying to explain is that the ESXi and NSX have completely bypassed in an SR-IOV operation. And SR-IOV is essentially blinding the application using the poll mode drivers in the virtual machine to a [ Pacific server ] or the NIC hardware combination, essentially making every benefit of virtualization impossible to achieve. So you cannot have HA, DRS or vMotion in this case in SR-IOV. Whereas if you look at enhanced data path, it is actually a single VM driver uses VMXNET3, and it abstracts the way individual NIC vendors from the VM driver, right? So here, there's an abstraction mode. And as a result of that, we are able to achieve and expedite the vMotion HA and DRS and the [ NIC teaming ] functions that you are aware of, that are available today. So decouples the VMs from hardware and enables virtualization benefits. So this is highlighting the advantages and disadvantages of SR-IOV over enhanced data path. But there are pros and cons, right? And at the same time, what I see is that SR-IOV has latent imitation where as an enhanced data path is high on core consumption but still able to offer higher throughput. Let's go under the weeds of what happens in the DPU. So in a DPU, we have 2 methods of communication from an acceleration perspective and optimization as well at the same time. So we have the VM direct path and then we have the emulated mode, which is the default mode. So what happens is that we can have the emulated mode, which uses less CPU versus the offloaded DVS. There is no guest memory reservation, there's no driver dependency performance accelerated by the hardware in the DPUs in this case and offers higher workload density. If we then look at the VM direct path, so it offers near 0 CPU consumption because all the services are concentrated into the DPU accelerator from a hardware perspective, offers complete guest memory reservation. It requires, however, the VMXNET version 7 driver and offers high performance and low latency data path model. So the other important bit I wanted to talk about is the distributed firewall and advanced threat prevention, which is ATP, part of the NSX ecosystem. So NSX provides East West security control and function at every vNIC level. So what really happens is that at a virtual NIC level, we have this concept of slots. So the slots exist without an NSX ecosystem when we introduce NSX binaries into the hypervisor in this case, DPU. Those provide additional functionalities from a Layer 4 to Layer 7 straight full firewall where we can intercept traffic and provide black or white listing at scale. So what that means is that every single vNIC that exists in the enterprise has a distributed firewall element attached to it. And in this case of distributed -- in this case of DPU, we are offloading those services into the DPU accelerators. And what that means is that we are able to scale at the same way we were able to scale the original theme and context of NSX. So if we need more firewall capability, what that means is that we are able to add another host and scale at the same time. So basically, it is scale out to break out concept from a distributed firewall perspective. So it's one single firewall to manage, and it's scaling at an enterprise level. So what we can then provide is we can provide enhanced security posture and performance with DPU. So it's an evolving architecture for modern Zero-trust security. So what that means is that if there was any CPU utilization that was happening inside the x86 platform is now offloaded to DPU and you have that CPU resource utilization at hand to deploy more virtual machines or container ecosystem in the same host. It provides more performance as well at the same time, as you can imagine, and we've discussed it right now. And this has not been tested in production right now. Again, remember the disclaimer and improves the security [ cost ] and visibility as well. And this is mainly talking about IDS/IPS as it stands today. So it provides comprehensive observability for all traffic ecosystem because we are able to pull that telemetry at the SmartNIC or the DPU level. And NSX Manager can pull in all of that limit for you to provide deeper level of forensics should we need to. And that is the huge value add from that perspective. So in summary, the key capabilities exist, and they have been [indiscernible]. So as a result of announcement of NSX-T 4.01, those were announced and NSX 4.1 The feature functionalities available from a distributed firewall perspective. So subject to you have a DPU ready node, you can actually utilize the NSX services offloaded into the DPU. So a quick recap on benefits and summary. So preserving all operations and have availability features for NSX and vSphere, single API UI for operational models. So there's no change despite of the fact that you are running 2 copies of ESXi, one on the DPU and one on the same x86 hardware. You have a single management plan to expedite those functions, standardize on VMXNET3 and paravirtualized DPDK pole model driver. So we have standardized on that as well that we discussed earlier around the architecture and 2 modes that are available and we have 0 CPU utilization. So these are those benefits. So enabling distributor services engine, what is [indiscernible]. So basically, we have the ESXi unified installation to install ESXi on Host and DPU simultaneously then enable offload via VDS. And even at the, a, which is the point A here in the slide, the ESXi would detect the presence of DPU base SmartNIC and give your ability that you want to enable offload services via the VDS. And at Stage C, we have unified NSX installation as well or we can use it, the life cycle manager to do that. And then we can configure the VMs for the 2 types of modes of VM direct path, UPTv2 or the unified mode in this case. So distributed services engine then extend the virtual infrastructure into a distributed control fabric. It provides single secure and clear and easy to use operating model across different type of workloads and provides isolation and air gap resegregating it from operating system and offloads infrastructure services function to the DPUs. So a bit on pricing and packaging, so it is available in vSphere 8 today. So if you're already a vSphere 8 customer or vSphere+ customer. So depending on the type of license you have, you may be entitled to use the DSE functions? Same for NSX. So if you are a sector enterprise plus customer, it gives you the NSX offloading capability from that perspective. A little bit more information on additional resources, so you can visit vmware.com/DPU to learn more about the DPU and the evolution in terms of the offering. There is also a good YouTube link that talks about [indiscernible] distributor service engine end-to-end demo. And then the DPU-based acceleration NSX deep dive where it goes through the inception of VDS and configuration and NSX manager. On that note, thank you so much for listening to me. And I'll now hand it over to NVIDIA, who will talk about their positioning and the joint collaboration with VMware And Dell. Thank you.

Thank you to VMware for presenting on distributed services engine. My name is Seth Henneman. I am a Solutions Architect for NVIDIA, specifically focused on Bluefield and the integration in distributed services engine. so I'm going to talk today about some more detail around BlueField-2, some of the performance and efficiency savings that we've realized by some of the performance testing that we have done with vSphere and NSX distributed services engine and then introduce our NVIDIA launchpad platform for customers and partners to get the hands on. NVIDIA BlueField-2 supports acceleration and offloading of distributed services engine features through the use of the entire platform. So BlueField-2 utilizes the NVIDIA ConnectX-6 technology and then on top that, it adds eight 64-bit, A72 arm cores, and those arm cores are used to run ESXi on the DPU itself. In addition to that, there's series of hardware accelerators on the DPU. Some of these are used for specific NSX and vSphere services that are accelerated and/or offloaded. I specifically want to talk about the SDN offloading that happens as part of distributed services engine. So there is an embedded switch on the BlueField-2 that utilizes NVIDIA's [ ESOP square ] technology, and it allows us to hardware accelerate data path processing, which is pivotal in the implementation of the uniform pass-through feature of distributed services engine. So we could program hardware flow pipelines within that embedded switch. And it allows us to accelerate the data path of network traffic through the hypervisor to the guest virtual machines. Meanwhile, the control plane for that traffic is still handled by hypervisor services, thereby allowing customers and operations to still utilize vSphere's hypervisor services such as vMotion and DRS. This is something that was not possible prior to UPT being implemented on the BlueField-2 due to the direct PCI binding of the guest virtual machines to the underlying hardware. How does the BlueField-2 hardware acceleration and offloading really translate into performance improvements for vSphere 8 and distributed services engine? Well, NVIDIA in partnership with VMware and Dell created a white paper and recently released this report outlining some of the performance benchmarkings that we've completed to measure performance and efficiency savings. We're going to talk a little bit about some of those results. But before I do, I'd like to encourage listeners of the webinar to access the full paper linked at the bottom of the slide. The first thing I'd like to talk about is the test itself. This is a Redis benchmarking test. Redis is an in-memory key store. We use a Mem Tier to drive these performance benchmarking numbers. These charts highlight different configurations that were used with environment, the different bars are different colors, represent those different configurations. So the light gray is a traditional link, which is a non-BlueField running overlay traffic within this performance testing. The dark gray is where we're utilizing DPU offload. However, we're not utilizing the full acceleration of uniform pass-through. We are utilizing the default mode. The green line is where we are utilizing UPT or uniform pass-through, on the BlueField-2. So this is full acceleration through vSphere and NSX. The first graph here on the left, and I'm specifically going to talk about the 80 server instance results, which is all the way on the right of that graph. At scale, the users can see that the traditional NIC, 25 gig NIC performed at about half its capacity, slightly above that for overlay traffic when scaled up to 80 Redis servers. In contrast, uniform pass-through when turned on, on the 25 gig BlueFields that were utilized in this test allowed us to get much closer to line or a performance for the cards at roughly 22 to 23 gig. So a dramatic improvement in just general throughput of the card by utilizing the BlueField offload. Similarly, from a latency perspective, if we again look at the 80 instances of Redis, we can see the latency is dramatically decreased when moving from that light gray bar or the traditional tick all the way to the BlueField offloaded and accelerated UPT implementation. So again, a dramatic improvement in latency and bandwidth for this testing benchmark. Similarly to the performance improvement reported in the last year, we're looking at CPU savings. Now for the 80 instance test all the way on the right of this chart, we can see that in the case of traditional NIC, not utilizing any DPU offload, the performance tests were utilizing 15 of the core on the X86 host. When moving to embedded mode, half of those cores were offloaded to the DPU and freed up on the x86 host. And then lastly, when going to full acceleration with UPT implemented on the BlueField-2. We were able to free up all 15 of those host cores and all the way to 0 as far as the core usage on the x86. Lastly, from a power efficiency perspective. Moving from the traditional, NIC, to the BlueField-2 with full acceleration, we could see that our efficiency saving numbers increase at those full scale of 80 Redis instances as well referenced on the right side of this chart. So as you can see, the watts per million transactions drops from roughly 45% down to closer to the low 30% when moving from traditional NIC to BlueField-2. So across all of these tests, again, we're seeing performance improvements from a bandwidth and latency perspective, we're seeing significant CPU savings, and we're seeing significant power efficiency savings. This is all again highlighted in the full white paper that's linked at the bottom of the slide. And again, I encourage users to access that and read it. One of the most commonly asked about use cases for offload and acceleration of BlueFields is NSX security services. One such security service I'd like to highlight here is distributed firewall. Users of vSphere 8 distributed services engine, BlueField-2 can offload NSX distributed firewall policies and rule sets to BlueField in order to free up x86 host CPU cycles. Traditionally, traffic would pass through a SmartNIC and the distributed firewall policies would be applied using CPU core [indiscernible] x86. This would not only eat up CPU cycles, but also limit the bandwidth that those -- that could come into those x86 hosts through the traditional mix. By offloading and accelerating those policies of BlueField-2, you can dramatically increase the bandwidth coming into your guest virtual machines to the point where it can reach almost line rate for the BlueField cards with a distributor viral rule set that is enterprised scale. Customer or partner that's interested in getting hands-on experience with BlueField-2 distributed services engine, NVIDIA provides access to our launch ad platform and specifically to the vSphere on BlueField DPU launchpad labs. These are accessible through the link that's on this slide or you can go to Google and look up NVIDIA launchpad and look for the infrastructure tab and then under there, you will see vSphere on BlueField DPU as well. These labs will give you free hands on with vSphere 8, NSX and the BlueField-2 distributed services that you're running on the BlueField-2. This will allow you to get hands on experience before making any purchasing decisions. And so we encourage you to do that as well as watch the fireside chat, where the CEOs of VMware and NVIDIA talk about the partnership with -- between the 2 companies and with Dell. To go into a little bit more detail for the labs I referenced on the last slide. We have 2 launchpad labs available for users to get hands on with. The first one is our administrator persona lab. In this lab, the users will get hands on with NSX, learn how to configure a distributed switch that's offloaded to the BlueField-2 and then learn how to create virtual machines with UPT-enabled on them and then also utilize hypervisor services on those VMs like vMotion. So this is our operations lab, where users will get more comfortable with the environment. The second lab then is our developer persona lab. This is where users on an already configured environment will get hands on with some of the Redis testing on a smaller scale that we highlighted in the white paper. So they'll be able to see some of the performance improvements for themselves on these hands-on labs. So again, I encourage you to access these labs and sign up if you're interested. And our NVIDIA account managers will be in touch and schedule you for time on the clusters. Now I'm going to hand it off to Dell to talk more about their product lines that support NVIDIA BlueField-2 and VMware distributed services engine [indiscernible]. Thank you.

Good morning, and good afternoon, everyone. Thanks Seth and Atif for setting the stage and sharing amazing content around Project Monterey. I'm Sohail Akram, VMware product manager, and I'm responsible to derive Dell and VMware joint solutions. It is really, really very exciting time and we at Dell are proud that we are part of this journey since beginning when started under the umbrella of Dell Technology in 2020. Now I would like to highlight the value proposition of Dell VMware and NVIDIA joint solutions. So we, as Dell VMware and NVIDIA can simplify and help our customers toward the next step in their journey of digital transformation with our joint solutions. We, as Dell, have a long-standing relationship with VMware and NVIDIA, which spanned over more than 22 years. And over these years, we have worked together on thousands of joint projects and offered our customers very reliable and high-performing solutions. So we have started working to gather on this project from the beginning in 2020 and our engineers have spent hundreds of hours to get there with VMware and NVIDIA. This collaboration, of course, efforts include incorporating respective technologies into this integrated joint solutions. Now why it matters to you? Of course, the highlight of this why is that the joint solution for you is ease of doing business and also the expertise in deployment of your required solution from Dell. And once you have this joint solution sourced and deployed within your new environment, then we at Dell will offer you a unified support across both hardware and software stack. We, at Dell, we have the industry most VMware certified professionals who can offer you help on both hardware and the software stack. And of course, [indiscernible] industry made the importance of OMEVV which is OpenManage Enterprise for VMware vCenter plug-in, which helps to bridge the physical and virtual gap by automating and consolidating the key PowerEdge server management cost in the VMware vCenter consul. And of course, it also offers sort of life cycle management alongside [indiscernible] feature. So as Dell, we are in a unique position as we are the first and the only vendor right now to offer NVIDIA DPU-based solution in the market. So, at Dell, can offer our customers diversified options to deploy NVIDIA DPU-based solution to meet your needs. If you're a customer currently running future architecture and you want to deploy this DPU solution within the same environment, then you can have Dell DPU-supported PowerEdge server. And if you are running an infrastructure based on VMware vSAN HCI solution, then Dell is the only vendor in the market who can offer you a jointly engineered HCI solution based on free [indiscernible]. And if you are planning to deploy it using your existing vSAN solution-based environment of vSAN Ready Node then we have a good news that we are going to offer soon DPU supported vSAN Ready Nodes. So regardless of your existing environment, you are able to sold a joint solutions to meet your existing environment needs from Dell. I have already shared with you what options you can get from Dell. So let me explain what you're going to get in this solution. So you can buy these DPUs off the shelf, but you will not get the seamless management as you would get from the joint solutions. This solution minimize complexity by managing the DPU. The DPU infact, simplifies the DPU management by treating a DPU as peripherals, which means high-speed management connection internal to the server instead of external like other digital solutions. Host OS and DPU OS is kept in sync automatically. And when the host OS is updated, iDRAC from Dell pushes the same version down to the DPU. All the services on the DPU are provided by Dell and users do not have to load their own services. And least, but not the last because we integrate the DPU, you will be able to use your existing tools such as vSphere, NSX Manager, iDRAC and OEM EEV or VxRail Manager with this solution. So currently, we are offering DPU Solutions only as factories to offer coverage, and you cannot buy Dell badge and NVIDIA DPU as kit because the server needs specific [ adviser ] configuration as well as factory install of internal management cable. We are working with our product group and expecting soon the availability of Dell badge and NVIDIA DPU to help those customers who looking to utilize the investment on the existing infrastructure. So what you need to buy, [indiscernible] offer. So currently, R750 and R60 are 2 Intel-based models, which are certified for DPU and you can buy them with NVIDIA BlueField. Right now, it's only single DPU support from VMware, but expecting soon releases of -- updates of vSphere 8, we will have a dual DPU. iDRAC 9 enterprise or data center is required with this. And also vSphere 8 enterprise plus, 8.0 license is required to use edge of the monetaric abilities. The customer has to be on the vSphere 8 version. So of course, we also offer optional deployment services. And all of these solutions will be provided with Dell L1, L2 support [ fund ] for both Dell hardware and software stack. And then, of course, you will also need, if you want to take down advantage full blown services than NSX-T for [ .01 ] enterprise plus is the license required for that. What we have on the road map. Currently, we only offer Intel-based 2 models. But in the coming days, soon, we will have 16 G-based Intel and AMD PowerEdge servers as well as going to offer NVIDIA 100 GB DPU as well. So VX will also have only factory-installed DPU supported option. And currently, we are not offering kit availability for VxRail. So those customers who want to utilize their existing VxRail-based infrastructure, at this moment we don't have a solution for that. So when it comes to VxRail systems, they will be shipped with the DPU hardware installed and ESXi preloaded and configured on the DPU. For day 1 perspective, automated cluster deployment and set up of ESXi host and DPU that will be available. Guided workflow of host and DPU networking configuration for day 2. Of course, ESXi update for DPU integrated into VxRail LCM as well as for DPU driver and firm [indiscernible] integration. We will also offer guided workflows for node addition and removal and replacement of DPUs and system with DPUs. It will offer the seamless and automated operational experience through native integration with the vCenter, automated nondisruptive upgrades, patches, node addition of retirement with intelligent life-cycle management. And hence, VxRail is going to offer you unified and consistent manageability with a single secure operating model. What models we have to offer. So currently, we have VxRail E660F, P670N and V670F from nodes perspective from VxRail. And all of these will be provided -- will be supported with NVIDIA 25-gig BlueField-2. In the future, we're expecting the selected models based on [indiscernible] platform, PowerEdge platforms. They will be offered as well as NVIDIA 100-gig DPU. So contraction for everyone on the call. If you're a customer engaged with your Dell, NVIDIA, or VMware accountee or engage with your channel partner for more information. And if you're a partner attending this seminar, then identify your potential customer for DPU and engage with your Dell team. Thank you very much.

Thank you, everyone, for your time on this seminar. I hope this webinar will help you to get your next step in your journey of transformation. Thanks, my colleague from VMware and NVIDIA, and have a nice and lovely day.