SNIA Experts on Data
Storage Futures – It’s All About the Data
Dr. J Metz, Chair SNIA Board of Directors, and Scott Shadley, Chair SNIA Communications Steering Committee, discuss the association’s plans for the coming year, which include work on SNIA SDXI Smart Data Acceleration Interface, computational storage, DNA data storage, SNIA Swordfish®, SNIA Emerald™ and sustainability. They also provide a fascinating insight into how they see the industry developing over time.
For additional information on the technologies mentioned, check out the SNIA Educational Library.
SNIA is an industry organization that develops global standards and delivers vendor-neutral education on technologies related to data. In these interviews, SNIA experts on data cover a wide range of topics on both established and emerging technologies.
About SNIA:
Dr. J Metz, Chair SNIA Board of Directors, and Scott Shadley, Chair SNIA Communications Steering Committee, discuss the association’s plans for the coming year, which include work on SNIA SDXI Smart Data Acceleration Interface, computational storage, DNA data storage, SNIA Swordfish®, SNIA Emerald™ and sustainability. They also provide a fascinating insight into how they see the industry developing over time.
For additional information on the technologies mentioned, check out the SNIA Educational Library.
SNIA is an industry organization that develops global standards and delivers vendor-neutral education on technologies related to data. In these interviews, SNIA experts on data cover a wide range of topics on both established and emerging technologies.
About SNIA:
Welcome to the SNIA on Data podcast. Each episode highlights key technologies related to handling and optimizing data. This is SNIA on Data podcast storage future. It's all about the data. Philip Alsop of Digitalization world speaks with SNIA Board members, d J Metz and Scott Shadley on SNIA technical work and where they see the industry developing over time.
Philip Alsop:So I'm going to chat today just to catch up with SNIA, and see what goes on, and the first place or the obvious place, I guess is you've recently elected a new Board, so I'm interested when that happens. s it all changed or complete, or is it very much business as usual? But some new people, so a few new ideas. How does, a new Board have any impact on SNIA, or not, as the case may be?
Dr. J Metz:Oh, anytime you bring in somebody new, they're always going to change the dynamic just a bit. Every year we have an election and so we kind of split down the middle in terms of which companies are up for election and which are which are not, because every every term is for two years and there are some exceptions for appointments and that kind of stuff, but in general we have an election for every every two years. I mean actually had quite a bit of change over in the last couple of years because of retirements and people moving on from different companies and that kind of stuff. So every October we have we have the elections, and in this year we've had a couple of changes.
Dr. J Metz:By and large, we actually have one of the most technical boards that we I've ever been engaged with, as a matter of fact, people who have had quite a bit of experience in the standards world, both inside and outside of SNIA. It's a very close knit kind of community of people who work on standards and so you kind of see the same people over and over again and we're very fortunate to have a number of people inside of inside of the SNIA board who bring a lot of experience, a lot of enthusiasm, a lot of passion and just some serious brain power, and not to say that we haven't had that before. But you know, I think one of the things that is kind of interesting is that there's a bit of a synthesis that's happened over the last couple of years. That make it very encouraging.
Philip Alsop:Okay, and then, in terms of the sort of program for the year, I know you've got various initiatives and ongoing initiatives, so it would be good to maybe have a bit of an update on each of those and the place to start us. Well, what I've chosen is the smart data accelerator in phase two, s. So it'd be good to understand sort of where we are and you know, if we were to chat at the end of the year where you'd hope to be with it.
Dr. J Metz:Well, so the smart data accelerator interface, the SDXI, is a really kind of interesting development. We've had a lot of companies that have been involved in the development over the last couple of years. So let me explain exactly what it is so that people have a good understanding of why it winds up being important. So it's called a data mover technology or data mover engine, but it's a software base. So the reason why that becomes important is that if you're going to be doing direct memory access, typically what you have to do is it has to be some sort of hardware enabled ability to move data from one memory location to another memory location, which means it's very specific to the hardware. So what a software based data mover allows you to do is have privileged software. Secure privileged software have the authority to move data from one memory location to another memory location, independent of the hardware type. So AMD, a these are the kinds of hardware chip manufacturers that are working on it but then sort of the systems integrators like Dell the SDXI TWG Chair happens to belong to Dell but then the software, like the virtualization companies and the OS companies. They're interested because by writing to a single software based standard they can port it from one place to another and they don't have to worry too much about well, they always have to worry about compatibility but you get a little bit more flexibility by being able to have it portable. In a software based environment. There's still room for hardware based data movers, don't get me wrong. But what this allows us to do is some very interesting and cool things that have knock-on effects. So let me give an example of what it would be used for.
Dr. J Metz:If you ever take a look at an application that's running in a container and you just kind of go down all the turtles all the way down to the bottom, you realize there's a lot of layers of abstraction, a lot of layers of abstraction. So if you want to do something like zero out the memory or if you want to move data from one application memory space into a storage VM or something along those lines, the Pachinko chip has to go down a number of layers of abstraction until you finally get to the hardware to do the movement and then has to go all the way back up to the application. So what this allows you to do is say I don't have to go through each and every one of these abstraction layers. Basically, what the application can do is the application can use the interface SDXI to talk to the hardware to do the movement from one memory location to another. And that's when you can do a lot of accelerated data movement from one memory location to another. And in virtualized storage environments this is particularly important because you do a lot of this kind of back and forth, back and forth, and this is one way of just bypassing all those abstraction layers and that's just one way of doing things. By the way, that's just one possible use case.
Dr. J Metz:There's a whole bunch of stuff that are currently being done to explore what you can possibly do in CXL fabrics, for instance, and smartnecks and computational storage devices, because each and every one of these has a memory location that has to move to another memory location. And so what? Right now SDXI is Intra server. It is inside of a particular node, but always using talking with a kind of a processor. But that's not a permanent thing. It's going to be expanded in the future to do quite a bit of flexibility. So does that make sense? Was that a good question?
Philip Alsop:That makes a great sense, and what sort of activities? Or, by the end of the year, what would you have hoped to have, moved on with this? Where would you hope to be?
Dr. J Metz:Well fulfilling the roadmap goals that people were trying to do and making it a bit more robust. But, as with any standard, of course, once it gets finished, once it gets technically stable, then you wind up having the products that become available. It doesn't really work too well to have it the other way around, since the specification can change. So we're expecting to see products on the market that have some support for this maybe limited first, but eventually what's going to wind up happening is that as these things start to become more prevalent and the use cases, we're going to see a bit of an evolution.
Dr. J Metz:Our goals are going to be talking about security improvements. There's security built into the process. You cannot run this without having a specific security protocol that's already been initiated. So the assumption is that once you undo the instantiation, it's privileged software based upon the security protocols that are already being followed. But that could still be improved. So there's some of those things that need to be done. So one of the other things that we're looking to try to include are quality of service features, some latency improvements, performance improvements. Right now it's kind of an all-in-nothing catch-all for DataMover, so now we're going to try and fine-tune it quite a bit in version 1.1 and beyond.
Philip Alsop:Okay, and in the answer you referenced computational storage and I know that's another sort of big area that you focus on. So again, be good, just to have a bit of an update, just to know where you're. What's going on with that?
Scott Shadley:Yeah, absolutely. So that's where I'll jump in. So that's been a near and dear to my heart for a few years now. Really, what's going on there is last year, in 2022, we finalized and released the 1.0 of us of an architectural spec for the ways of implementing in the building blocks required to implement a computational storage device, and we have a couple of different of those based on the spec. The most recent thing that we're working on in the working group, and it kind of the next step, is the software API that's associated to that so you can actually implement it with the software layer, similar to what SDXI is doing for memory.
Scott Shadley:We're doing that software implementation for kind of computational storage. One of the uniqueness of this particular situation is we are building an architecture that is somewhat agnostic to the protocol, because we can't say that it's just going to be one device versus another device or one way of communicating and another, and so we actually have a significant effort in partnerships and alliances that we do on behalf of SNIA working with the working groups like NVMe. So the first NVMe command sets for computational storage are being developed on that side and, to Jay's earlier comment, we're kind of all sitting in these different groups and different aspects and different roles and so by way of having an alliance with those guys, we can make sure that the architecture that we've defined as being implemented in such way that the command sets align appropriately and things like that. So a lot of the focus for 23 and computational storage is finish up the software work a little bit more on the security side of things. To Jay's point, security is a huge thing right now that we are working on across the board of SNEA and we have a SNIA t on security that's been very, very busy in 22 and looks to be in 23.
Scott Shadley:And then from a you know what's after that in the computational storage it's work on use cases defining, you know, getting customer feedback. So we've got some customers that have deployed early prototypes from various different vendors and getting them to come in and present what they've done with it so that others can see. Because one of the unique challenges that always exists with a new technology wing in the industry is the early adopter versus the fast follower. And sometimes it takes a little while to get that early adopter to present what they've done and then people go, oh, now I can do that, and so we expect to see the computational storage market officially take off, if you will, in 23 24. There were signs that it was going to do it in the 20 20 time frame. The pandemic sure didn't help with that, but it gave us time to finalize some of the specs and work that we've been doing on the standard side, so that was actually beneficial.
Philip Alsop:And I know another area that's quite exciting to have a look at is the DNA data storage. So what I say, what is is on the agenda for that in the next sort of 12 months, or what sort of works being looked at?
Dr. J Metz:And data storage is a really interesting thing and in a lot of ways, the propeller spins the other way when it comes to this, because it is it is radically different than any kind of traditional storage than we've played with before. So you know, the very concept of being able to store data in DNA sounds like science fiction, quite frankly. I mean, it's the kind of stuff that you'd see in the sci fi movies from the 90s and 2000s and and the. But when you start to think about the practical applications of it, you start to realize that there's a lot of coordination that needs to go on. So, for example, how do you encode the data in DNA? Well, that's a major question that needs to be resolved. How do you read the data once it's encoded? How do you make sure that my data in a in a DNA string, is portable to a number of different places? All of these different questions are really having to be started over from the ground up. We we don't have the same kind of legacy approach to solving these problems, mostly because the the math gets pretty complicated pretty quickly and compared to what we have done in the past by moving from tape to disk, to non volatile memory Right? So what the DNA group is doing is that they are. They have a couple of different subgroups, right? So one is what they're calling the archive Rosetta stone, the DNA archive Rosetta stone, or D a R, s, and effectively it's. It's what.
Dr. J Metz:What I was talking about? It's the ability to define the way that the data is read. You know, how do you make sure that you know what you read is what you think you're reading, and so on and so forth. How do you code the data? Like a? The way that this was described to me is that the encoding is sort of like a master root record, right? So how do you read the MBR version of DNA so that it becomes understandable and not just in a proprietary fashion? And these are the kinds of things that happen, and by being able to stand on this, sanitize on this, developers and vendors can save a lot of time and money and also the words to market to be able to get the kind of a ubiquitous approach to, to storing and retrieving data.
Dr. J Metz:One of the other things that they're looking to do is is figure out how to do scale, right? So when you're talking about zettabytes of encoded data inside of DNA, and then you have to have systems in place to be able to to manage it and track it. So there's, you know there's. Like I said, you have to start off at the ground zero. You have to, you have to identify what your criteria is for success are, what your criteria for failure is. How do you do metrics? How do you do capacity? What's the transfer rate going to look like and what's expected? Those are the kinds of things that are being done in the technology roadmap area Long term, long term retention, you know what happens over time.
Dr. J Metz:So that means that you have to do some environmental conditions and tests that are not necessarily completely unique to DNA. You know, we do, we do environmental condition testing and all kinds of storage development, but with DNA, we don't necessarily have the guarantee of foresight, knowledge, and so that's why we have to make sure that you, you have to test the technology, you have to test the methods and those environmental conditions over time, or one of the big initiatives that are currently being done. I think one of the things that they're looking to do by the end of the year is you know, come up with some agreed upon, you know, structure and architecture, with those three different components in place and once they do that then they can start getting into some of the, you know, some of the protocol areas.
Philip Alsop:Okay, and in terms of its potential, I mean I was going to ask later on, but I'll ask it now. I mean it's been a while since, if you like, solid state was the major leap forward for the storage industry. Is DNA storage the next sort of big thing when the issues you've described, or is that not the case? Is there something else out there, or is DNA storage the you know got a big future?
Dr. J Metz:Well, it seems like there's a number of different companies and people who are looking at DNA storage. Probably you may recall, you know, a couple decades ago the holographic storage was a big rage, right, but it didn't never really quite reach critical mass. I think that probably DNA storage right now, even outside of SNEA, is getting interest that may rapidly approach the critical mass, as necessary for pushing it over that edge, you know, to get the momentum going. I think it probably has the most potential for being a. It is radically outside of the box, thinking in terms of how we thought about long term storage for a while, and I think that there's probably enough companies to make it worthwhile and happen. Nobody, of course, can predict the future, but I think it's probably got the best opportunity and potential.
Scott Shadley:And to that point, you know, thinking of it, it does really fit into the long term storage. So if you, you know, do any looking into it today especially, it's not a fast read and write type of architecture, right, it is definitely to define around more of an archival type of situation. So, you know, does it take out tape who knows? Right, but it is definitely, to Jay's point, something that we're very excited about having and helping drive from a perspective.
Philip Alsop:Yeah, although I guess anyone who says tape is dead will, is you know, pushing their luck on, because that's been said so many times before. Exactly.
Dr. J Metz:Yeah, so let me put that to bed real quick, right? So seismic data on the planet, which exists of hundreds of exabytes, is all on tape. It is recorded directly from the earth through the ships into tape. All of it is recorded on tape. Exabytes and exabytes, hundreds of exabytes of data are all done on the seismic data and it changes. Right, the earth changes. So tape is never going to go away if we're only that one use case. You know, the use of tape goes so far beyond what happens in a data center that would be silly to think that, an arrogant to think that any because somebody moves to a different technology that's better for them, that it's not going to be good for anybody else. Tape is a ubiquitous and extremely important technology for many industries.
Philip Alsop:Yeah, and then, in terms of the next one on the list, I've got the, the Sniers Swordfish initiative. So again, it would be good to have a bit of an update and sort of outlook as to what works going to be undertaken on that.
Scott Shadley:Yeah, the swordfish and the, you know, the governing body of that, if you will, is our SMI or storage management initiative, which supports our, you know, storage twig or storage management twig. So we have a lot of work going on there. If you think of it from a definition of Snier perspective, we have technical working groups which do they in you know, in the trenches type of work around things like the sort sort fish documentation and implementation. And then you have our SMI, which is our initiative, which is more in the marketing and push out for one of you. So from that point of view we have a lot of things going on in the Alliance space. So, kind of back to that. You know how do we work well with others in this space.
Scott Shadley:A lot of the focus on the technical work group today is to continue the evolution of swordfish as the primary storage management interface and scalable interface, if you will, by way of working with the DMTF and the redfish side of things and working within VME and what they're doing in the way of their management work. That they're doing because there's new MI specs coming from that group, and so a lot of the focus for the technical work group today is now we've got this really good baseline set. It's making sure that the baseline can continue to grow appropriately across the infrastructure and the entire ecosystem, because that's kind of a primary focus on that. A lot of it's around, you know, collaborative work with the other working groups as well, as the XI, for example, will have some aspect of management involved as it evolves into a larger footprint, and so that's a lot of the focus there and it's still, you know, to some extent educational too. So there's a lot of focus that's going to be on how do we get more people interested in driving this forward. It's, you know it's become the new baseline. We've got new labs set up someplace where people can come in and test and put in hardware and do things like that around those types of things.
Scott Shadley:So we're very excited about what's going on with Swordfish and the associated working groups. You know NVMe, dmtf, ofa, you know pick your letter acronym alliance that we could be involved with. We want to make sure that that's been done appropriate across the board. As far as adoption goes, we've seen a lot more effort on that front. We've got lots of traceability of what's going on in the way of people that are deploying and using this interface a lot of support from a lot of the member companies and their customers as well, so it's doing great things. Right now we're seeing, you know, a nice breadth of opportunity for it to continue to grow and evolve and come out with the next iteration of that. You know additional specs and additional features that it can support over time.
Dr. J Metz:And it's, you know, eminently modular right.
Dr. J Metz:So a lot of the stuff that's going on inside of the Swordfish interface is applicable to a number of different technologies, which include, you know, cxl.
Dr. J Metz:You know that includes, obviously, nvme, which is a big part of it, but it's not just those technologies. I mean, the one of the things that has a particularly big advantage over the previous management structures is that it really, now that the infrastructure is in place, adding an additional functionality, features and environments is not the forklift upgrade that people, you know, may have had to do in the past. I think one of the things that you know may actually be missing is that some people have gotten so used to being locked into a particular type of management metaphor that they don't really have the imagination to think that they could possibly do things in a relatively easy fashion by simply plugging into, you know, this architecture. It's a shame, really, because it could make their lives an awful lot easier. But, as Scott was talking about, you know, people are slowly starting to realize that there's usefulness for them and it can expand their you know their uses.
Philip Alsop:Okay, and I was going to ask later on about your work with other partner organisations, associations, etc. But as you've referenced them there, it would be good to understand that. As you say that I think you've named a few, and there's the Ethernet Alliance and the OCP initiative it would just be good to understand a little bit of the various collaborations you have going and what you've outlined some but other things that you're looking at with all the partners.
Dr. J Metz:So Sneha, go ahead if you want to go.
Scott Shadley:I was just going to say it actually parlays really well, because SMI represents one of the good examples of how that's come together really really well. The last two years, for example, at several of the large industry events like Super Compute, sneha drove what we call an open standards pavilion, so we secure a space at these events. It's a large footprint instead of just a small little one off booth and each of these industry associations are able to come in and take part side by side and you can literally you know hey, I was talking to someone about this and you look to your neighbour and there's the other Alliance member that can talk to what's going on in that group and so you know to your point it's OFA, uci, pci, sig, ieee, DMTF. Our goal with that is we do actually have a strategic Alliance focused group, a committee within board members and members, that manages all of these relationships. Some of them are very open relationships where deep down technical work is shared real time. Others are collaborative, where you know current specs are shared in a pre-production state so that we can make sure that they're aligning appropriately.
Scott Shadley:But the overall just is to make sure that there's synergies as we move everything forward and different groups have different focuses on aspects of things. So you can look at that as a way of, you know, trying to be the weaver of all of those together in some fashions, kind of where Sneha's been able to be a great success because you've got open compute, for example, is driving a lot of new things. But if you look at the OCP spec for SSDs or for storage today, it references Sneha, it references DMTF, it references all these other groups specs as part of its you know specifications. So it's a reference doc for everybody else. And so working together with them to make sure that we don't, you know, break one while doing another in these groups is very important and that's a key for us. And so we do have a lot of work being done in there. Across all of these different organizations Soda, ofa, you know you could pretty much pick one and Sneha somehow in way, shape or form engaged with them to make sure that we're collaborating in the marketplace.
Dr. J Metz:Yeah, we're kind of in a unique situation because, unlike some of these other organizations, sneha is both a standards development organization and a trade organization, and so I think that, depending upon which avenue you come into Sneha from, it's easy to assume that that's, that's the only thing that Sneha does. One of the things that we're looking to do over the next couple of years in particular is to strengthen these relationships on both the development side and the trade side, because Sneha has the already has the format down for for being able to help those related organizations and you know they can help us too. It's a mutually beneficial environment, as we're looking to do.
Dr. J Metz:I think in the past what's been happening is that there there was some potential concern about who's going to do what.
Dr. J Metz:You know, there's a bit of a hesitation on turf and that kind of stuff, but, as I said before earlier on a lot, it's a very small organization, a small environment of people who do standards and and it's easy, once you get into it, to say, okay, this is, this is something that should be done in MDM Express, or this is something that should be done in OFA, this is something that can be done in Sneha, and it's not that difficult to say okay, we can actually work together to do this.
Dr. J Metz:Our job right now is to facilitate those environments and those relationships so that you can move from one to the other and put the right place, you know, use the right tool for the right job, and so it's a, it's a key focus as long as I've been chair, to get to the point where you know the alliances that we have with industry bodies is strong and healthy, and I'm glad to say that we're actually making quite a bit of progress on that. We've got we've got a few, you know, a few in the works and we're strengthening the ones that we've got.
Scott Shadley:Yeah, and I mean, if you look back at the comment about DNA storage, what started as the DNA storage alliance? It was a group of companies that were focused solely on that effort and as they looked at how do we take this to the next step, they approached me and said you guys have the infrastructure around how to put work groups together, you have the infrastructure on how to manage IP relationships, you've got all these relationships with the other groups and so the data the DNA data storage alliance is now a technology affiliate umbrella underneath SNEAD. From that very example and you know smaller things that are starting up like that are great for us to take advantage of larger organizations like NVM Express or SOTA or whatnot, or perfect partners in crime.
Philip Alsop:And then perhaps the last one, at least I think that I'm aware of, is the SNEAD, the Emerald Program. So again, it'd be good to understand what's on the agenda, you know, over the next 12 months or so with that.
Scott Shadley:Yeah, this one's actually very interesting. It's near and dear to heart of our chair emeritus, mr Wayne. He has been driving that for a very long time and it's interesting that how relevant it's becoming now as people are looking at carbon neutrality and credits and things like that around sustainability. All of that kind of gets umbrellaed, if you will, to some extent through the work being done by the Emerald Program, the Green Storage Initiative, the Green Storage Twig and those types of things and the trick with those again come back to some more industry work. Right, so we can do a lot in these organizations to drive awareness, attention.
Scott Shadley:But sharing that and getting involved outside of the standards groups, even to some extent into the federal and governmental standard groups, whether it be in AMIA or in the US with, like, the EPA those are a lot of the work that's being done in the Green Storage work and the Emerald work to drive initiatives and focus on how to make storage better or make aspects of storage, interaction with storage, frameworks of storage, networking or otherwise more compatible to what we need to do in the market, moving forward just as sustainability and green perspective, which is kind of why Emerald is a green gem.
Scott Shadley:That's why it's called Emerald because it was focused around that kind of effort to begin with, and there's a lot of work being done around how to manage and interact with those types of things. So take, for example, kind of one of the things that you can look at is the form factors of SSDs are changing, the power envelopes are getting larger while we're running into a power crisis and a heating and cooling situation. So this is where the Green Storage Twig under the you know guises of all the work being done there and the umbrella under Emerald can help drive. How do we make these systems more efficient while driving more power or more performance out of the systems and things like that. So it's not just you know, if I plug it in the law, which is a consumer, how do I reduce it? It's literally looking at bits and pieces of the ecosystem and how can we drive better work there from that perspective.
Philip Alsop:And I'm just wondering as well I mean more generally what responsibility. Obviously you've got the initiative, but how responsible or not do you think SNEERS should be for driving that sustainability agenda? Obviously, individual companies have their technologies and they're aware of the pressures around environmental issues. But do you think you should be a sort of cheerleader for Green Storage and try to persuade members in the industry generally to take it more seriously? Or is it up to individual companies to sort of do their own thing? Just your thoughts on that.
Scott Shadley:Yeah, I think one of the interesting things. To your point, we're doing a good job of helping drive that, not only for you know, our particular members, but for the end customer platform. So, as I mentioned, for example, epa came out with data storage, data center storage, version 2.0 test that was based on Emerald V4. I had to read it off a slide because it's a, you know, a tongue twister, but that's where people that produce products go out to grab a test suite, download it and upload it to the EPA, showing how well they fit the future of sustainability and power and performance type of thing. So the groundwork that we've laid within SNEERS and using the Emerald platform actually kind of puts us forefront on that and the ability to do it, because the relationships exist with the different industry associations In the US EPA.
Scott Shadley:There's a whole bunch of different aspects of that that come into play from, you know, things around TGG and lot nine regulation updates and those types of things, all these different non-industry but more, you know, governmental, federal, whatever type of aspect. You want to look at it bigger picture, not just the component level, vendors and things like that, but actually things that mean more to everyone than just the members, if you will. So I we're very excited about what they've been doing, continue to do and we'll continue to drive as we move forward around the sustainability side of things.
Philip Alsop:And then maybe, just as we sort of finishing, it'd be good to understand if I know it's a big question so if you're able to sort of give an answer, summarized, but in terms of how you'd assess the state of the citizens. As I say, it's been a while, if you like, since SSD sort of arrived and everyone got excited, and I'm not saying things aren't happening, but sort of just sort of now the profile is perhaps a little bit lower at the moment. All that change and what are the sort of challenges we mentioned? The sustainability, one of the other challenges, whether the cost of living crisis etc. Supply chains and also on the plus side of the opportunities to grow the space. So yeah, it's a big question, but if you're able to give me some thoughts, I'd be grateful.
Dr. J Metz:It's a fair question. I think one of the things that history has taught us is that there's sort of an accordion effect and expansion and contraction. You see, like, for example, in 2016, 2017, you saw a major influx of cash into the virtual, into the VC rounds for different source startups, and that will happen again eventually. Right now, we're kind of in more of a contraction, kind of a moment, less investment. Obviously, the global economy has a major role in that, as it always does.
Dr. J Metz:From a technological perspective, one of the things that is kind of interesting and I think that when it starts to expand again, I don't think it's going to expand to where it was. I think what's going to happen is that technologically, we're working on a collision course between storage and memory. So what persistent memory has allowed us to do is cut out a bit of the time necessary for moving data from one place to another. So the question really is not necessarily only what the different media is going to give you, because you can get more capacity, you can get more speeds and feeds, but what the time is going to give you is what people are going to capitalize on. So if I have a difference between 160 nanoseconds in memory and 8 milliseconds on a hard drive, that I cut down to 100 microseconds on an SSD, down to 1 microsecond with the persistent memory. What does that time give me? That's where the question really comes into play, and it's just outside of a load store budget. So what CXL is trying to have to, what CXL in particular is having to cope with as a technology, is that your margin for error shrinks down quite a bit as well. So if you've got some load store that has a 300 nanosecond budget and you're 500 or 600 or 700 nanoseconds away, what do you do? Well, you have to build some intelligence in the system of where these things wind up going.
Dr. J Metz:That's where the SDXI starts to play a bigger role, because now it fits into a place that was kind of awkward before, but now we're focusing on that awkward place to try to see where the balance is between where memory is, where storage is, where persistency is, where volatility is, and it all goes up the stack. It all happens to be one of those things where it all rolls uphill into the application space, because everybody wants more memory, everybody wants more speed. They don't really care how they get it. But that's why we do what we do is because we try to help figure out the best ways to get these things to use the tools where they're needed. Always use the right tools for the job, and time is one of those tools. That is kind of ethereal. You can't put money on the development time for time. You have to put it onto the objects that affect time and at the end of the day, that translation is where I think we're going to wind up going.
Scott Shadley:Yeah, and if I were to add into that, if you look at it from more of the architectural point of view, at a hardware level we had things like people were focused so much on Moore's Law, the doubling of density, halving the price type of thing, and whether it's dead or not. We were seeing alternatives to persistent memory. We've seen a shift in what's going on, with things like the Optane and that type of technology being moved out because of the lack of immediate adoption. But if you look at it from a global perspective, we had Moore's Law that was driving the NAN semiconductor in memory space, which is still there, but people don't talk about it as much anymore because it's less relevant to what is actually the bigger pieces, which ties back into Jay's comment on time, and that's our friend Mr Von Neumann who created this.
Scott Shadley:I have a CPU, I have a memory block and I have a storage block. Those are the three things you're going to work with in any compute system and we worked with those for years and years and years and decades. And now we've kind of broken that down and we've said you know what, it doesn't have to be fine lines. So now we've added things like the XPU, a DPU, a GPU, an IPU pick your favorite name of some form of an acceleration attachment between either memory and storage or memory and DPU, and then you push it the next step down, which is where we are with things like computational storage and even SDXI as it relates to the time movement. So when you look at how our industry is driving and you look at performance as a key metric, it's no longer how fast does each drive, it's how fast do a group of drives in an ecosystem work together to provide the results you're looking for in the time you want to get it done in, and it's also comes down to a power and cost perspective, and those are big ones too. There was an event a few years back that was a super computing event in Brazil that a friend of mine went to, and they had two competitions One, how fast can you get it done and how cheaply can you get it done at the best speed.
Scott Shadley:And those are very key, relevant things that a lot of the work being done with Insnia are driving around through, and to closure for those types of things, because if, for example, qas, quality of Service, qos whatever acronym or variant you want to say of that has become a paramount topic in a lot of the working groups that we have, and it's not how many millions of IOPS or how small a nanometer you're building, it's literally the aspect of time. And where can I save time and at what cost can I save that time? So that is definitely kind of the thing that we're looking at. So it's moving away from thinking of everything as a hardware and a software, but thinking about what you're working on, which is the data.
Scott Shadley:Right, everything is about the data. How do we take what data we gather to his point earlier about all the seismic data of the earth and turning it into something that we all find valuable. And that's really where we're starting to kind of shift our focus over. The next little bit, too, is not just what is the component, but how does that component impact the data and the work that needs to be done on that data to make the end result the most feasible and logical for the user.
Philip Alsop:Maybe just one final promise, because you've given me a lot of your time, for which I'm very grateful. The geopolitical situation Does that worry you at all? I appreciate we probably, as you know, as individuals can't do much about it, but do you worry that the storage industry is vulnerable? Do you see a change in the supply chains, as we're beginning to see in the semiconductor industry, for example? Just, does Sni have any role in terms of ensuring storage security in that terms, if you like, and what's your role in that scenario?
Dr. J Metz:Well, we, we, well, we're not a lobbying organization, but one of the things that we have decided that we're going to move towards is to be a little bit more assertive on positioning papers, right, what the consequences are, because a lot of people who are making those regulations and this is true all over the world, nobody, nobody, is immune from this. They have consequences and those consequences can have very profound deleterious effects if they're not understood, well understood. So, some of the regulations involving power, some of the regulations involving security, some of the regulations involving data, privacy, mobility you know, you know those kinds of things, reporting they have consequences and oftentimes they have very complex, intricate, web-like consequences. You know the ripples go pretty far out in those ponds once you start throwing pebbles at them. So what we're going to be doing you know what we're aiming to do in, you know, over the, you know the next few months, years, so on is to is to take our expertise and share it with you know those who are willing to. You know those who are willing to learn about it.
Dr. J Metz:Power is a good example of this the, you know the security and privacy stuff. You know. Recently, you know, we had a meeting internally on some of the things that are going on with privacy and the privacy laws and, quite frankly, they scared the living crap out of me, so I wasn't aware of it, and I think it's very important that we really hammer home to the general public, to the regulators, to the people who need to know that there are aspects to this that they may not be aware of, and we have that for good, bad or ugly, we have people who understand these and need to be able to share that, you know, for the betterment of everybody, and that's one of the things that we're going to be doing. We haven't had a chance to really, you know, to synthesize that yet, but it is a very strong goal for us to be a responsible member of the industry society, and that's one of the things that we're looking to do.
Scott Shadley:Yeah, and to kind of the the GOP's of it, if you will. We're all been in the industry quite a long time to see these cycles right. We have the ups, we have the downs. We have Coloma downturned whatever you want to turn me on to use for it but they've existed for 30 years and one of the unique things about them is they'll grow. We'll get a whole bunch of companies. There'll be consolidation as they start to shrink. We kind of have this contraction and expansion.
Scott Shadley:But one of the benefits of SNEA, for example, is if you look at the Rolodex of members, it's about the membership and who's participating.
Scott Shadley:That's helped drive that 25 years of history.
Scott Shadley:The companies that associate to those names change frequently to some extent or will change from time to time. J and myself have been part of this board under multiple company names, for example, as just a perfect way of looking at it. So SNEA has been able to keep the members and the talent and the individuals that can answer the questions that he was talking to through their careers, because we're not focused on you know where you work as much as what you're able to do and how you can contribute, and we welcome that effort and a lot of the work we do around gaining new membership isn't? You know, we're not looking for body count, we're looking for that expertise that can help drive the industry forward. And where do you work as a perk right? In what kind of space you're in, going from a storage company to a networking company to whatever, to whatever. Those individuals have helped drive that success and SNEA has had a very good bedrock of membership to help keep that happening, regardless of how these cycles have hit the marketplace.
Dr. J Metz:And everything is related to everything else. You know so to what Scott was talking about. You know the annoyement stuff. You know the computer is related to the network, the network is related to the storage, which the storage was related to the computer, and so on and so forth, and no one person can be an expert in all of it. And so it really is. You know a matter of one plus one equals three, because you know when you have one person who's an expert on another thing, you have another person who's an expert on another thing. Now you have a broader understanding of things than just those two people you know in and of themselves. By being able to work together, you create a third factor of the relationship between different concepts.
Dr. J Metz:And you know, from a geopolitical perspective, most people nowadays are looking for quick fixes. You know they're looking for either people to blame or they're looking for a button to push. Either way, you're going to wind up getting disappointed and if you're not careful, you're going to find yourself in a situation where you're going to choose the wrong button, you're going to blame the wrong person and you're going to be no better off than you were before. What we're looking to do is try to create a systematic approach to being able to answer some of these complex questions and make them consumable for people in a plain and understandable language. Yep.
Philip Alsop:Well, it's been fantastic to talk to both of you. I really appreciate, as I say, you're giving your time. So, jay and Scott, thank you very much indeed, thank you.
Scott Shadley:Absolutely. Thanks for inviting us. Thanks for the time. Yeah, appreciate it.
Kristin Hauser:Thank you for listening. For additional information on the material presented in this podcast, be sure and check out our educational library at sniaorg.