Storage Unpacked Podcast - Storage Unpacked 266 – Architectural Choices in Storage System Design

Episode Date: February 24, 2025

In this episode, Chris discusses the options available to storage system vendors when building modern storage appliances, with Bill Basinas, Senior Director, Product Marketing at Infinidat. The conversation derives from an observation on architectural choices, following the move to AMD processors from Intel for the latest G4 systems built by Infinidat. AMD offers a greater core count per processor compared to Intel, allowing Infinidat to move to single socket designs, while gaining improvements from PCIe 5.0 and DDR5 memory. Ultimately, this discussion highlights how modern storage system design can take standardised components and build flexible architectures, implementing most features in software. For Infinidat, that could mean expanding its range of solutions for smaller enterprise requirements, or building out products specifically for Edge use cases. Although Bill did not reveal any future plans, the implication is clear - watch this space for future evolution of the InfiniBox architecture to a wider and more varied set of hardwaree configurations. Elapsed Time: 00:37:13 Timeline 00:00:00 - Intros 00:01:15 - How do vendors choose the hardware components for storage systems? 00:02:30 - What are the main (storage) technology challenges for customers? 00:04:08 - Customers want predictable data features 00:05:55 - Capacity demand continues to grow relentlessly 00:07:30 - Infinidat features are built into software 00:09:35 - Most AI requirements wil run on existing performance storage 00:11:20 - Modern hardware provides significant flexibility for system design 00:15:00 - AMD gives access to single and high core-count processors 00:16:10 - PCIe 5.0 provides for faster SSDs and power efficiency 00:18:46 - Infinidat has introduced smaller form-factor solutions 00:21:32 - Multiple cores will always get used! 00:25:53 - Infinidat G4 architecture provides for in-place controller upgrades 00:28:22 - Storage arrays should become more “virtual” 00:34:10 - Data services implementations are very different between vendors 00:35:55 - Hybrid architecture still has value in the Infinidat world 00:36:20 - Wrap Up Related Podcasts & Blogs Storage Unpacked 258 - Introducing Infinidat G4, InfuzeOS 8 and InfiniSafe ACP #202 - Enterprise Storage Consolidation with Phil Bullinger from Infinidat Infinidat adds customer value with SSA Express and improved SSA capacity Copyright (c) 2016-2025 Unpacked Network. No reproduction or re-use without permission. Podcast episode #e4dr

This is Chris Evans and today I'm joined by Bill Bacinas from Infinidat. Bill, how are you doing? Hey Chris, great. Glad to be here. Hope all is well with you and the in the new year. Yeah, very well Thank you. Very good. And it's good to be able to have a chat with you. I was gonna say without Eric That sounds awful, doesn't it? I don't mean that Don't mean that in a nasty way by not having Eric here, but it's nice to talk to Another friendly face from Infinidat. But before we get into talking about what we're going to talk about, why don't you just tell people what you do for Infinidat and then we can get into it. Sure. Yeah. So you mentioned Eric and Eric and I go way back.

We've worked together a number of times, but here at Infinidat, I am the Senior Director of Product Marketing and I run all the marketing positioning and messaging aspects of what we do here at Infinidat. I've been here, oh, just crossed my three-year anniversary. And as a person with a gray beard, you've probably been in the industry a long time. Been around the block a few times, as we say, Chris. You know, the concrete on the sidewalks has worn a big groove from where I've walked over the decades.

Excellent. Excellent. Good. So this should be really an interesting conversation. Now let's give everybody some background as to where we got to in terms of why we're having this discussion. And this really sort of came about because you and I have talked about the upgrades you've done to your platforms recently in the last sort of six, 12 months, which is focused around moving to an AMD architecture and some of the changes.

And now one of the things that struck me about what that was, it would probably make an interesting discussion to talk about how you go about sort of making those decisions, how you look at the technology and how you sort of migrate towards, you know, picking a different architecture, but also how when you're designing your software, you build things that allow you to have

those architectural choices and to be able to make those changes. So I thought it'd be good to go over sort of some of the thinking, some of the background, some of the things that you're trying to do towards helping customers by doing this, and possibly a little bit about where it might sort of lead people to think

you might be headed in terms of the technology in the future. Now, you don't have to give away any spoilers, but I think it'd be interesting at least to have that sort of level of discussion. I think it's a good way of people understanding where Infinida is headed. That's what I'm thinking. Yeah, yeah.

That all sounds great. And I think we can provide you a lot of good information on kind of where we've come from, where we are today, and maybe a few peaks into the near future. Excellent. Right. So let's start with customer challenges. I mean, I think, you know, I look at it from my perspective and say this, but, you know, I think customers are always looking for the same thing. They're looking for something to be cheaper, faster, cheaper to operate, as well as cheaper to buy. And they want it to be efficient. But I think now we're moving in into an era where people want things to be so efficient that they can integrate it with automation and they want, I guess, consistent consistency.

And when they deploy something, they want to know that, you know, you're not putting 15 different software solutions in or hardware solutions in. All of those things, I think, have come to the fore and they've been there for years. But what are you seeing as the sort of the current demand of what customers are looking to achieve? Yeah, you know, you bring up a lot of great points, Chris, and they're all extremely valid. And depending on the customer you talk to, it always seems to be somewhat of that subset of different items. They want to effectively manage their infrastructures.

They want to make sure that their solutions are providing them a medium to long term, solid capability that they can really trust within the infrastructure of their business. Cost and efficiency are certainly top of mind, especially with the cost of data center environments today and the restrictions that they have because of growth. Organic growth and other areas that has put constraints into the data centers. They want to make sure that the solutions that they're putting in are as power efficient as possible, that they're, you know, in our

particular case, storage efficient as possible. But I'll tell you one thing that they're not willing to give up on is a consistent set of predictable data services, right? I'll tell you one thing that they're not willing to give up on is a consistent set of predictable data services. Right? I mean, you know, we've gone back quite a while, you know, quite a bit about our solutions, but the core value propositions, you know, that we bring to the table around reliability, 100% availability, you know, our cyber resilience guarantees with our InfiniiSafe technology, that's, you know,

that's a new area that's still up and coming when we talk about the cyber aspect of what people are really needing to consider, not only in their overall IT infrastructure, but specifically to their storage and data requirements. So absolutely, all those are definitely valid things and things that we talk about with customers every day. Yeah.

Now, from your perspective, to turn this on its head slightly, turning it the other way, as a vendor, you've clearly got challenges in the technology you deliver. And I'll give you one example. So imagine that we've seen the growth in storage that's been anywhere between,

you could say 30%, 100% growth per year. And in fact, with AI now, that number might be even higher with the ability for people to take data from sources that aren't human generated, like sensors and various other things. It could easily be 100%. And I think if we'd not had improvements in the technology, our data centers would be the size of, well, the size of cities, because we'd, you know, the floor space that would have been needed to do this stuff and the power requirements would have been impossible to manage.

So there have to have been improvements to manage the scale of the data. So from a vendor perspective, what are you looking at fixing in order to make sure you can deliver to what customers want? Yeah, so great questions because for the exact reasons that you talk about, the data proliferation and data growth aspects is not slowing down at all. We continue to expand our data center footprints by providing better overall capacity solutions to our customers in different levels. We're at the high end of the market, the? The enterprise level of the market. And some of the workloads that we go after

are consolidations of multiple workloads. So that's a big piece of it. If you think about wanting to shrink the data center footprint and optimize the data center, typically that means you have to do one of two things. You either have to separate your workloads or keep them separated, which continues to add to sprawl. And when you have sprawl,

that has its own challenges from a management perspective, how you lay out your data centers, the amount of equipment that you need to acclimate into those particular environments. But when we consolidate workloads, we're typically bringing together multiple workloads into a single storage architecture. This leads to some of the architectural things I think we'll probably end up getting to. But our software architecture and our performance characteristics give us a rock solid capability to provide that level of consolidation and thus bring

efficiency to the workloads and that those data centers and those customers are really looking to combine but adding an ease of use to it as well, right? Because we're eliminating that sprawl, we're providing a very simple and user-friendly set of interfaces, and our software, our Infuse OS, is really what provides the functionality here, not the hardware that we sit on that we leverage and take advantage of, but the software is really what gives us

all the performance and availability characteristics that give customers that rock-solid feeling, right, that we're a very valuable asset in their data center to manage their data. I think I like the idea of the discussion about consolidation because I think when we look at the current AI trend and gosh, everybody's going to talk about AI in one form or another. And if I was playing the old Dallas game where you're supposed to take a drink every time JR takes a drink, which is what we used to play in the 80s and the 90s. I think if I took

a drink every time I said something about AI, I'd be on the floor by now already today. But if you look at it, you've got different types of data, as you said, and you've got different uses for that data, maybe some for the AI systems, maybe some from somewhere else. And should you create lots of copies? Should you not? Should you try and consolidate? There's a real trade off there between operational issues and cost and all the rest of it. So consolidation definitely has a massive piece to play, I think, in the current industry. It does. And you know, our solutions kind of play across the board, right? You know, the funny part is, is we, I don't want to say sat back, but we watched the AI industry over the last 12 months specifically

To see kind of where it was gonna go a lot of people a lot of our customers were kicking the tires right there We're trying to figure it all out and how they deploy it within their infrastructures You know, there was a lot of noise out there, especially in the storage industry about needing to have specific built workload type support You know with GPU direct type interfaces and things like that to what are typically massively created AI clusters, right? In order to build these types of infrastructures. And the reality is, is unless you're, you know,

a government entity, a giant lab, or some kind of a massive entity, the vast majority of the needs to service the AI infrastructure, especially where we some kind of a massive entity, the vast majority of the needs to service the AI infrastructure, especially where we've kind of positioned ourselves, runs perfectly fine on a high, well-performing set of storage. And we have that and we're in that, we're kind of in that data center area where we don't have to do a whole lot of things special in order to provide those extended capabilities in there. And we're really focusing on

retrieval augmented generation, you know, and from an AI solution perspective because the training methods are, you know, are one thing but bringing truth to the result is hugely important. And that's really where RAG architectures are focused. And we maintain a lot of that data for customers. And it's a great fit for us to play in that architecture and that we retrieve a lot of generation architecture. Okay, great. So we'll maybe have a discussion about the architectures of storage for AI in another day. But let's go back to talking about today's discussion around

architectural choices in terms of hardware. Without a doubt, the industry has changed since when I first was involved in this. So when I first was involved in the industry, I'll put it this way, IBM was probably the only storage seller and there was no raid in those systems, which it tells you how long it is since I've been doing this. However, today we tend to use what I guess I've sort of termed as commodity hardware, which is probably, I'm not being unfair to it,

but it is standardized, shall we say, standardized hardware. The server form factor is pretty consistent, processes, memory, IE interfaces are fairly standard. So, but that always gets better over time. Every year you can look at it and say, we've gone DDR3, 4, 5, we've gone PCIe. Now typically 5, but 6 and 7 have already got the specs lined up.

So they're going to come along at some point. So you've got a lot of stuff in your arsenal that you can choose from in order to build modern systems. And you get a lot of flexibility in that choice. A whole lot. We've never been shy about talking about the fact that our systems, our hardware platforms, that we deliver our software on are commodity-based. We do not engineer controllers.

We do not engineer drive interfaces. We don't create our own drives like some others do, have decided in the industry to do. We use off-the- shelf enterprise class components in our solutions. So you're absolutely right. It gives us a lot of flexibility. And you know, you mentioned AMD and that we've made that shift from a dual socket Intel platform to a single socket AMD platform, right?

And we switch manufacturers of that platform as well. And you know, that's why we're a software company, then the power of what we do is in the software. And we can take the best platforms available and mold them into what we need and what our customers need, and not have to worry about taking long engineering kinds of cycles to create very specific hardware environments

to try to create the availability and performance and resiliency that we build into the software, which is unique. Yeah. So I follow, obviously, I follow what all the different vendors are doing. And you're right, there are some vendors that have gone down, the orange vendors that have gone down the route of custom SSDs, if you like. And they've done that for their reasons and that's fine. There's lots behind that. You've got a mixture of storage and hardware, sorry, software and hardware capabilities

which deliver your technology. And the interesting thing I think when you look at how vendors have brought those platforms to market is if you're organized, each generation that comes out and from an Intel perspective, it would be currently fourth generation Xeon, I think we're on now. A lot of the vendors, as soon as that comes out, they've got the opportunity to get in there straight away and develop

their platform for that next solution. And then it becomes a software question because as you said, you've got somebody doing the hardware for you, you take a platform and you build on it. I guess what I'm just, what I'm saying there is that, you know, your choices initially would be in lockstep with a vendor like Intel, but having AMD there, you've now got two choices. You've got multiple choices, but you don't have to be tied to one or the other. And you can, you know, you can be as flexible as you like. Absolutely right.

Just because we've made an initial, you know, the shift from Intel to AMD doesn't mean that we cannot still support Intel platform. We can spin those on a dime, right? If we have, you know, if we need to, and if we have to, or if in the future, you know, customers say, hey, we need this on an Intel platform because that's their strategic platform within their operations environments.

It gives us a lot of flexibility for exactly what you're saying. But when we look at some of the initial choices here when we went to AMD, it was a leapfrog type of a thing. It wasn't just an evolutionary type of a process when we made those considerations, Chris. It was really revolutionary because as you said, single socket, 64 core, highly efficient CPUs, DDR5 RAM, PCIe Gen5 infrastructures. Just from a hardware perspective, Chris, we obviously had to make some changes to the software to accommodate the compute and bus structures, right? But that was really about it. We didn't really have to change anything as far as data services go. And without even a whole lot of tuning,

because again, we're all about consistency and providing a very rock solid platform for our customers. So, right out of the shoot, we see a two, two and a half X increase in overall performance. We see better efficiency going from 48 cores to 64 cores, increasing that core count by a little over 30% and dropping to 64 cores, right? Increasing that core count by a little over 30% and dropping the power consumption, right? That and

another important consideration when you're trying to build a lot of more powerful infrastructure to keep the power envelope in the same line, right? So in essence, a 20% reduction in power utilization per core. So it's a win-win-win type of a scenario for us by being able to literally take an off-the-shelf server with that kind of capability. And that's not the end of the road, right? As I mentioned, where there's a lot of things

that we are still doing and looking at and implementing that are gonna continue to enhance the performance strictly from a software perspective by leveraging all the capabilities that are available in that platform. So some of the things I think I can see that you get the benefit there. Obviously, if you look at say PCIe5, it's double the throughput of PCIe4, slightly lower latency. And clearly, looking at the market, although you use a mixture of different media, we can see that some of the larger capacity SSDs were being constrained by the fact that they were on PCIe4 and their throughput

was definitely constrained. And as soon as you start to see the PCIe5 models, that restraint is lifted. But obviously, if you want to scale a system, you need to have higher scale infrastructure within there. But as you also pointed out, if you want to make something that might be slightly smaller or more consolidated, I know that PCIe5 is more power efficient than say PCIe4. So bit by bit, you've got the ability to both go up the stack for scale, but also potentially down the stack for scale. Absolutely right. You know, and the other things to consider too, especially as you shrink the footprint and we've done that, right?

I mean, with our G4 announcements this past spring, for Infinidat, who is typically a rack-based system, you know, a full rack-based data center type system, we announced a 14U system that can be installed in customer racks. That's a big departure for where we've been, right? And kind of gives us a glimpse into the future that we're going to continue to consolidate those. But we're not giving up on all our

enterprise feature sets, right? Because that's what our customers really come to us for. And they want to make sure that not only do they have a consistency in deployment options, because some of them have gone to pod architecture, some of them want to put them out into remote data centers or in manufacturing facilities, but they want that same ease of use, that same consistency and reliability, even though it's a smaller overall footprint. And again, the software architecture has allowed us to do that. And some of these new hardware architectures allow us to do that as well. And we're going to continue down that path. Yeah.

I think, you know, looking at, for example, some of the other things that are clearly obvious, the single core, I guess, allows you to put in a lower form factor server. So maybe, I don't know, I've talked my head, I don't know whether there were two of you before, but you know, you could see a scenario when you could move to a one-U server instead of a two-U server, or you could move to some of those fat twins where you've got two servers that sit side by side

in the same rack. So there are lots of flexibilities around changing that architecture when you've got flexibility around the hardware, as long as you can still deliver the software on that platform and keep the resiliency. I mean, there's the challenge. Absolutely.

And this is an audio podcast, but I'm smiling, right? So you're absolutely right. You audio podcast, but I'm smiling. Right. So, so you're absolutely right. You've hit the nail on the head. Those options are all on the table and all things that we're putting into, you know, to play for our architectures as we continue to move, you know, continue to move everything forward. It's a massive amount of flexibility.

And you know, our customers are already seeing benefit from the 1400 series and what that's bringing to their infrastructure without having to worry about, not to belittle anything, non-enterprise architectures. We derive our performance from our software architecture, from leveraging the efficiencies that we create, our caching mechanisms and are software-based raid mechanisms in our architectures where we're not really reliant on the backend. We can do a lot performance-wise and things like that just strictly from that caching mechanism and data efficiency from a layout perspective that benefits us in even small

form factors. Okay. So I'd just like to ask you a bit of a technical question then, because here's something that I'm looking back at where we were, say, oh, 10 years ago in the industry. And there was a lot of talk around the idea that we would be able to use many, many cores in our processes and that having more cores was better because we could put some of them doing compression, we could put some of them doing this task and so there was a little bit of a rush towards putting in many, many core processors into these systems and obviously dual socket and all the rest of it. Having said that, we seem to have headed

down the route where we've got a lot more advanced extensions, instruction extensions to x86. And I reference in our notes here AVX 512 as an example. And I wonder whether, bear in mind you've gone to single socket, whether the core count is as essential as it used to be, or whether there's benefit from the instruction set helping you to do some of this work in the background. You know, I guess that's a good conversation to have with some of our architects. But from the bits I know about it and the facts of what it presents, I think it gives not only us but others in the industry that ability to utilize those cores in very specific

workloads too. Right? I mean, some of the AVX 512 stuff is very well oriented towards AI workloads, right? The vectorization processes and so on and so forth. Does it give us capabilities that maybe we can take advantage of when the timing's right? Absolutely. And the nice part is that they're not always on, right? So you're not kind of sitting there with resources being used in the CPU that you might not be

able to take advantage of, right? The fact that you can kind of switch these things on or off and write code specifically to them are hugely interesting. Whether there's a huge difference between a single core architecture, how it affects that or a dual core, I don't really know. It's kind of beyond my pay grade to understand how that how that those particular architectures work. But I do think that more cores, we're finding ways to take advantage of them. Right? There's no lack for utilizing the infrastructure that the CPUs and the servers are giving us. And it's really a matter of putting those to the best uses, right? That's what we're

always looking for, Chris. We're trying to figure out what's the best use to provide our customers, going back to the first steps we talked about, that efficiency, that cost efficiency capability, and that storage infrastructure capability that's going to carry them for the long term. We have very long term relationships, and our storage sits on a data center floor for quite a while with our customers. It's, you know, it's kind of interesting.

Phil Bolentaro, CEO says, you date servers, but you marry storage. That's one way to look at it. Yeah, okay. Yeah, that's, that's probably true. And I do think as well being in the industry a long time, you certainly don't want to change storage systems at the drop of a hat, because at the end of the day, as an example, networking is a very advanced game of, I think in the US you call it hot potato. So basically, networking is literally, oh, it's a packet, I want to get rid of it. And you know, you turn a network switch off and you put it back on again, you don't really lose much because the network doesn't really have a lot of state in it in terms of the data. Similarly, and the server, you know, server, you shut down, you boot up and the same thing applies. Storage is where your long term persistence is going

to sit and you have to be really careful. We have to be really careful. Absolutely. So you don't want to lose that. So ultimately, yes, the hardware is going to be reliable, and that's great. And it's good that you can use these different form factors. But interestingly, just going back and touching on it again, you highlight how the storage side of it is, sorry, the software side of it is so important to you.

Was it really that much of a change to adopt a different architecture? No. From the storage perspective, none are all crisp. Absolutely none. Right. Maybe some path and optimizations and things like that. Right.

But, but when it really comes down to that back end, we have enough of, if you will, I don't know if it's the perfect term, right, but we have enough abstraction in how we address the hardware where that's not as much of a concern, right? We certainly don't want it. We certainly wanna take advantage of the areas that we can, but overall, that's not a big lift for us for it to make that kind of a change

and have to be overly concerned about what's going on at the backend, right? I mean, all of our offerings stayed from a backend device perspective, Chris, as we moved to the G4, stayed exactly the same. Nothing changed. Okay. So let's put that in context then of customers who decide that they want to keep the data consistent. They want to grow over time. They look at it and they say, I want to stay with Infinidat for the next 10 years, 15 years, maybe could be longer than that.

It's not going to be the same box every year. It's not going to be the same hardware. They're going to want to be able to replace that over time, but they don't necessarily want to A effect performance B have to do mass scale migrations. So what I'm trying to get to here is the software is driving the ability to enable you to swap that hardware. And as long as you've got that software defined on consistent hardware, swapping boxes out in order to do replacement shouldn't be a big challenge.

It shouldn't be. And I think we proved that point, right? A couple of different ways. First and foremost, historically, we have had a fantastic, I think one of the best stories and not only stories, but capabilities in the industry when it comes to migration. And our in-family migration is 100% seamless. We have built into our software that there's no additional charge, right?

We basically include everything in Infuse OS. We don't charge for any of the core features and any of the data services, it's all inclusive. And one of those pieces is an inherent capability to do a fully active, active storage system, synchronous storage system, which a lot of our vendors, you know, charge a lot of money for and or have trouble kind of doing that efficiently. But we use that in conjunction with what we built as what we call Adam, which is our migration tools that give our customers

a 100% seamless migration capability. And that's been historically what they've done. Now this past spring with the G4 architectures, we decided that we were gonna move into a secondary option that we've made available that we announced what we call our Mobius program, our Infiniiverse Mobius program. And that's an in-place controller upgrade program where people who will start with the G4 platform

will be able to do controller-based upgrades in the future. Okay. So that means that if I buy, let's pick an example, if I buy the 14U system or I buy an entire rack-based system, you're just going to come in and pull the server and put an upgraded one in place? by an entire rack based system, you're just going to come in and pull the server out and put an upgraded one in place? Yeah. What we'll do is that whatever that next generation is, and we're looking at what that will be. We just launched the G4 this past spring.

And once that's available, we'll make that available and we're going to be very transparent about it. Our customers will know what they're going to get, what it entails, what the performance benefits will be, you know, and so on. And we're not tying it to service and support programs like some of our competitors do, right? Where you're kind of prepaying for that controller upgrade and really not knowing what you're going to get

at the end of that three or four years. So we're taking a very transparent approach to it. We'll make that next architecture available. They'll be able to swap out those boxes seamlessly to what they currently have and be able to move to that next generation. And if they decide that, hey, you know, the backend storage architectures are something that work well with what they have today, they can stay with those. If they want to move or shift to maybe new or start back in architectures that will, you know, may have in the future. That's great too. And we can do that completely seamlessly just with our native tools.

Okay. Do you know the, you probably never heard of a TV series called Only Fools and Horses? I would imagine. Can't say I have. Okay. It's very famous TV series in the UK for about 30 years. And within there, there was a character called Trigger who wasn't very clever. He's a bit unaware of what was going on a lot of the time. And there's a really funny story

about something called Trigger's broom where basically it's a bit like, I think you might have called it Washington's axe or Newton's axe where it's the same axe but it's had its head changed five times and its handle changed four times. Or Trigger has a broom that he uses to sweep the streets, which has had four new handles and two new heads.

It's theoretically, he thinks it's the same broom. I do think storage is headed in the same way where potentially a storage array can still be that air quotes the same storage array it was, you know, 10 years before, but actually in reality it isn't because if you can come in, pull out the servers and replace those over time. If you can come in and change the storage, physical storage over time, as far as the customer's concerned, they're seeing that same consistent application unit, you know, the storage array, but in reality it's not because it's constantly being upgraded and refreshed. the storage array, but in reality, it's not because it's constantly being upgraded and refreshed.

Yeah, absolutely. Yeah, again, I think you're right. We have a little bit of a different challenge, right? Because of our architecture being a three-way, fully redundant, triple redundant architecture, we're the first enterprise class storage system to announce a controller-based upgrade program. You know, all the other competitors that, true enterprise competitors that deal in that high-end space don't offer that. The controller upgrade programs, you know, historically to this point have all been offered by those mid-range type

dual controller type architectures, where typically one controller is somewhat passive anyways, right, in the Alua models and so on, which kind of inherently make that architecture a little bit easier to deal with. But we have, again, we have very resilient architecture and we didn't have to do a whole lot of new stuff besides test the processes in order to assure that we can very seamlessly do those kinds of upgrades for our customers in the future.

Yeah. And I say that as really positive because ultimately I can imagine, okay, so AMD has the step up on Intel at the moment in terms of things like PCIe 5 and the number of PCIe lanes it offers on its architecture. But those things never stay the same. There's always a, as you talked about earlier, a leapfrog, you know, one generation will come along and do something faster stay the same. There's always, as you talked about earlier, a leapfrog, you know, one generation will come along and do something faster than the other.

And I'm not gonna put any words in your mouth, but there are other hardware architectures that come along all the time. And we've seen ARM gaining some significant steps in various different places. There's even RISC-V out there as well. And not that those might be suitable at the current time for storage,

but you could imagine an entire suite of different things for different scenarios that take edge use cases, even Intel or AMD processes that are specifically designed for lightweight, lower powered environments where you might choose to put something like that in. Yeah, right now that's not on the radar screen, right? But I mean, that's a very valid conversation. Both you and I follow a lot of storage groups

and a lot of storage conversation and that's in the conversation, right? People are starting to ask questions about it. So I think at some point it probably plays itself out in one way or the other. But I think you're exactly right. Whether we're talking about the IoT from a couple of years ago, now a transition to more of the edge story, there's a whole lot of different needs that are going to need to

be met by a whole lot of different architectures. But we always know, Chris, right? Again, we've been around the block more than a few times. It's historically been hard to maintain functional consistency across those types of things when they've happened, right? And that's gonna be key, not just providing that level of functionality, maybe using those different architectures,

but making sure that it's 100% consistent from an operational functional perspective that provides an easy integration into a customer's extended data center, right? Or data or data set structures, if you will, across their enterprises. Yeah, it's interesting actually, Bill, with all of that. As you said, some of those comments, I'm thinking actually, yeah, one of the things that we possibly don't think about is the fact that if you go back, well, actually I've got a diagram on it, which I might put in the show notes on it. I may not, but I've got a diagram where I look back to when I started in the industry and a data center was physically a location where all of the equipment was locked away

and the networking was basically point to point across S&A. You know, it was very, very controlled, very walled garden. Yeah, and then we moved to a situation where things opened up a bit more. And in the last sort of 10 or 15 years, we've ended up with the data center concept meaning nothing anymore. You know, that we have a virtual data center almost

where customer data could be at the edge, it could be in the cloud, it could be on physical premises, it could be all over the place. And it's very difficult to sort of tap, put that down. So, so you as a vendor are gonna have to come up with some solutions in the future that meet all those requirements.

So, you know, you can, you can choose to tell me that or not. I guess you don't want to ruin Eric's Eric will probably get very unhappy if you give away any secrets. Yeah, but I'm sure you must have plans for the coming year. We have a lot of plans for the coming year, Chris. We have an extremely aggressive roadmap, right? As with all companies, we've gone through our end of the year projects and processes, and we've aligned our roadmaps for what our and our customer needs are for the coming year. We have a lot of exciting things that will come both in the area of the hardware architectures

of the products continuing to provide different levels of flexibility and the software as far as data services and things like that go, right? There's, you know, a lot of people think today that we've kind of reached the saturation level and we've reached this area of quality and what a lot of storage systems provide. I don't necessarily agree, right, in all cases,

because replication isn't replication, you know, and the way that file services are implemented, you know, are not necessarily the same, you know, our, you know, our competence, some of our competition implements their file services it's drastically differently than we do, right, and we, you know, we get the same exact performance on file systems that we do in block. So that's the kinds of architectural decisions and things that we do and the kinds of extensions that we want to make and keeping with that core value proposition of making sure that everything that we do is enterprise ready, enterprise class, regardless of the size of the footprint of where it goes. Because that's what people want to depend on when they're dealing with what they're fitted at,

in particular in our architectures and what they've learned to gain. We've brought simplicity to the enterprise storage environment. Typically, in the past, we both know when you said enterprise, it typically meant pretty complex. And the fact that we're doing an effective capacity scale on a single storage rack with our hybrid systems up to 17 petabytes and continuing to grow and look at the right architectures for our SSAs, our solid state boxes, which have performed unbelievably well since we've really put a big emphasis on them and now taking those architectures down.

The hybrid architectures, I think, are squarely up and still in that full rack environment. People who want those hybrid architectures want a whole bunch of storage. But in those solid-state architectures, they want small footprints, medium footprints, and large footprints as they need them in their environments. And we're going to continue to focus on all of those elements as we go through this year and beyond. Wonderful. And we're going to continue to focus on all those elements as we go through this year and beyond. Wonderful. And I think you've highlighted a very interesting point there, Bill, and that's that ultimately customers want that flexibility. Nobody deploys one thing in the data center anymore. 30 years ago, we might have put a mainframe in, it might

have just been one big box that ran everything. But the requirements now are highly diversified because of that requirement to meet all sorts of geographies and all sorts of other things that age. So you as a vendor are gonna have an interesting time producing products that meet all of those requirements. And it's gonna be fun to watch it for another 12 months, coming up in 2025, which I'm looking forward to. So let's see what happens and come back and well, you can come back, Eric can come

back and tell me exactly, you know, what you've been working on when you when you're ready to announce it. But for now, thanks for your time. It's been really good to catch up with you and I look forward to finding out what you're going to be working on. Well, we'll stay in touch, Chris. And as always, thank you for having us. And I look forward to continuing these kinds of conversations as we go through the rest of the year. OK, thanks, Bill.