SNIA Experts on Data

Breaking Barriers: Transport Technologies and Storage Innovation

SNIA Episode 17

Explore the collaborative efforts and expertise of SNIA in advancing data transport technologies. Key topics include the rise of NVMe, innovations like CXL, Fibre Channel's ongoing significance, and the role of standards in creating efficient data solutions. Hear from industry experts, Bill Martin, Craig Carlson, Fred Knight, and Anthony Constantine as they discuss:

• SNIA's role in data transport technologies and collaboration with industry partners
 • Key trends in SAS, NVMe, and Fibre Channel technologies
 • The impact of CXL on memory efficiency and architecture
 • The significance of UltraEthernet and its future applications
 • Understanding SFF and its importance in evolving storage solutions
 • The role of STA and industry partnerships in setting standards
 • How SNIA Swordfish has emerged as a key player in managing storage within the Redfish ecosystem  

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:

Speaker 1:

Welcome to the SNEA Experts on Data podcast. Each episode highlights key technologies related to data.

Speaker 2:

Welcome everybody to the SNEA Experts on Data podcast. My name is Eric Wright. I'm the co-founder of GTM Delta and the host of the SNEA Experts on Data podcast, very excited to be able to bring an amazing team of folks together here. We've really looked at SNEA in some of the previous chats, about some of the specific topic areas and innovation areas and research that's going on. This is a real great chance for us to go back and what's the core of SNEA and we thought about what are the different areas in which we look at SNEA in how data and being data-centric in a storage mindset and storage-centric in a data mindset when we look at how data works in the SNEA context.

Speaker 2:

If you go to the SNEA website, you'll see lots of great updated information on the different practice areas. Some will line up with some of the technical working groups and we'll talk a lot about that today. So I'm not going to pre-podcast the podcast because I got the greatest folks on the mic with me. With that, I'm going to do a quick ask of some introductions along the way. So if we want to get started, just have a quick intro about yourself for folks that are new to you, bill, if you want to kick us off.

Speaker 3:

Great. Thank you, Eric, and I'm Bill Martin. I am co-chair of the SNEA Technical Council and I work for Samsung as my day job and kind of facilitating this overall podcast with these experts in these various areas.

Speaker 4:

Excellent and Craig as well, Hi, I'm Craig Carlson. I'm on the SNEA Technical Council and my day job is I work for AMD. I do storage and networking. I've been involved in storage and networking Standards for over 20 years, including Fiber Channel and VME and many others.

Speaker 2:

All right, team FC. I'm old school, I love that stuff. Fred, you're next up.

Speaker 5:

Thank you, Eric. My name is Fred Knight. I work for Kioxia and I'm part of the SNEA Technical Council. I've been in that role for a number of years and have oh, I hate to admit it 30 to 40 years of experience in the industry and, going back to some of the early fiber, channel up through the latest and greatest NVMe technologies.

Speaker 2:

We all have this amazing thing where we used to cut off part of our leg and count the rings, and there's a lot of time and technology amongst those years, so I love it. But looking to the youngest of the crew and last but certainly not least, anthony, you want to do a quick intro for the folks.

Speaker 6:

Sure, and everyone's making me feel so young right now. So this is great. My name is Anthony Constantine. I'm an engineer at Intel. Anthony Constantine, I'm an engineer at Intel, and what I do for Senea is I'm on the technical council as well as I'm the co-chair of the SFF technical affiliate.

Speaker 2:

Excellent. Well, thank you everybody, and today we're going to talk about transport. Transport is an interesting section of the overall SeneA practice areas and I think the best way to lead it off let's talk about what does it mean when we talk about SNEA being experts on transported data? Bill, let's have you kick us off.

Speaker 3:

Great Thanks, Eric. So what is it that SNEA is experts on transport? What does it that SNEA is experts on transports? What does it really mean? So what it means is we work together with a lot of different organizations within the industry who develop transport standards. We try to bring some of that together. We try to provide education, Part of what SNEA has. We have a SNEA forum called STAY, or the SCSI Trade Association that is now part of SNEA.

Speaker 3:

Forum called STAY, or the SCSI Trade Association that is now part of SNEA. They do all of the what does the market need for SCSI? And they're looking forward to SCSI. They're arranging plug posts. We'll get into a lot of that a little more later in the podcast. And then we also have, as Anthony mentioned, he is co-chair of the SFF Technology Affiliate, where we develop standards for storage devices in terms of connectors and form factors. And then, finally, we have a large number of alliance partners. Number of alliance partners. Among those alliance partners in the area of transporting data are FCIA, NVME, CXL, as well as having liaison arrangements with all of the Insights T committees, which have recently been renamed by Insights to now be the ATA committee, the SCSI committee and the Fiber Channel committee, although I'm not getting the Insights names right for those.

Speaker 2:

Well, I think it's a really great kickoff for us too, bill, because if we look at how SNEA in general works, it really is. There's what work we do internally across different sections of SNEA practice areas, and then the alliance partners, and we're seeing stuff like CXL and some of the external partnerships that are coming in. So that's one of the things I've really enjoyed about SNEA's approach is that not just defining standards internally but looking to the industry and helping to give guidance, and then we're seeing more and more partnership come up. And when we talk about transports of the day, let's start by going through and I said Team FC. I'm going to go old school, Craig Fred let's talk about what are some of the important transports that are in play today, and let's talk about Fibre Channel and SAS and where that's it, where it is and even potentially where it's going well, fibre Channel probably would be considered one of the original and the oldest SAN storage area network technology.

Speaker 4:

the the Fibre Channel specification goes back 30 years and I was part of it for a good portion of that probably longer than I want to admit. I chaired the protocol community and a lot of the other work groups. The thing about Fibre Channel is that it has stuck around despite the fact that people every few years declare it dead. It's still in there. It's still being deployed in enterprise data centers. It's still a very important technology for a lot of the enterprise customers who are using Fiber Channel, basically because it's been around for a long time. It's reliable, stable, fast and they trust it.

Speaker 2:

We should do a dedicated one. Just called Fiber Channel is dead. Long live Fiber Channel right.

Speaker 4:

Exactly.

Speaker 2:

Now on the SaaS side, fred what's happening and what's happened, and where do we see SaaS being really prevalent today?

Speaker 5:

Now on the SaaS side Fred, what's happening and what's happened and where do we see SaaS being really prevalent today? Well, saas came about from the original parallel SCSI days as part of a smaller, more local fiber channel area and using a serial protocol rather than the parallel technology. But it has still continued, gone through a number of speed bumps, getting faster, adding new capabilities, improving the performance, still offering all of the full features that SCSI offers, all of the capabilities that enterprise environments have come to depend on. So it's still a significant area of storage devices, both SSDs and hard disk drives, so it's something that still is important today.

Speaker 2:

And this is always I used to well, I still say all the time is you know, you call it legacy, I call it production Quite often.

Speaker 2:

We get sometimes caught up in the hype of things that are coming and forgetting how broad and widely adopted a lot of these technologies are. They're still going to be in play for a long time, are they're still going to be in play for a long time? And as fast as technology moves and workloads change AI workloads being the hottest thing lately, of course what we're seeing is that more and more innovation in other areas allows these technologies to continue to add value and actually come to the fore. So what we're going to see, even on the software side, database technologies, which were very limited before now, are suddenly moving back to the front because of capabilities that are happening with some of the stuff that's going on the protocol side to allow software to kind of latch on to better capabilities and use more of what's there. With that, I guess we'll talk about the real hot topic, and the most prevalent I think that we see is NVMe. So, fred, let's talk about where that came from and, again, what's coming up in the NVMe world.

Speaker 5:

So NVMe is a technology targeting the solid-state drives, ssds. It's a protocol that's built specifically to achieve the maximum performance that that technology offers. So its primary transport technology is the PCI bus and the different speeds that that is available in different generations to get as much performance as possible out of those SSD drives, out of those SSD drives. So in addition, that was expanded into the network world with NVMe-OF, where the transport can be any number of different things. We have the ability to use RDMA, the ability to use fiber channel as the transport to carry the NVMe protocols. We have several others using TCP IP that can carry that as well. So the fabric technologies are still a very large part of the NVMe environment. In fact, second only to the PCI area is the fiber channel area. That that is a very large adoption area for the NVMe protocol. So we have it both in the direct connect technology of PCI as well as the fabric technologies everywhere, from TCP through fiber channel, through RDMA or many other kinds of wires that can carry those protocols.

Speaker 2:

Now I guess the next one. When we talk about the distributed workloads and the hyperscalers and what we're seeing with more disaggregated infrastructure. This launched also the need for what came to be CXL infrastructure. We're seeing new innovation and ideation coming around ultra-Ethernet. So, craig, do you want to share what is CXL at its core and UEC, and what are you seeing that's being done in the SNEA community related to both of those things?

Speaker 4:

Well, cxl is a new technology. It's both a bus and a network. Potentially, as the CXL matures, it's becoming more and more. It's reaching out more and more to allow for broader configurations. The one thing that really is an interesting promise of CXL is to use memory more efficiently, to be able to have pools of memory. You might have tiers of memory where you have your local host memory, which is your fastest access, and then the idea is you have CXL pool, which is maybe a bigger pool but it will be a little bit slower access because, of course, latency going out. You always have higher latency when you go out on the machine, but you still have a higher speed memory out there than you might have for other storage technologies.

Speaker 4:

And then UltraEthernet is the new kid in the block. It's taking a lot of the lessons learned over 20, 30 years in networking and applying to a whole new set of features for congestion, very, very large deployments in terms of millions of nodes, data loss protection and the idea is to deploy this in HPC as well as AI workloads and there have been existing technology that can do this PCB data set of bridging and Ethernet is one of them, and you would see where it takes that to the next step and uses the most recent technologies.

Speaker 2:

There's definitely a lot of stuff and we're going to have independent podcasts and some of the conversation. There's information that you can find as well. So I recommend folks that are dabbling with looking ahead look for a lot more. That's going to come up with what SNEA and the UEC folks are doing. So pretty cool there. Now, since we're talking about down to the form factor and we've got these wild new deployment patterns and operational patterns, we're changing how we're physically deploying. So let's talk about SFF, my BFF, anthony. What does SFF do for us and what has SNE sort of created with the work that you and the team are doing?

Speaker 6:

Yeah, so. So Fred and Craig talked earlier about the transport layer. What we're trying to do is basically build the, the electromechanical interconnect between, you know, a host and a device, and so what SFF does essentially? We create those cables, those connectors, form factors, all of those bits and pieces that are needed so you can physically assemble these devices together to go talk to each other.

Speaker 2:

And one thing that popped up was EDSFF, because we never have enough acronyms in our world, so tell us more.

Speaker 6:

Yeah.

Speaker 6:

So EDSFF which I guess a word of warning, never have engineers design a name for a device but EDSFF was built as a ground-up solution to basically solve a problem around existing storage form factors.

Speaker 6:

A lot of the form factors for storage were built around the hard disk drive and as we went into the SSD era I mean not that our hard disk drives aren't still around, but going into the SSD era we needed new form factors to be able to cool them, to be able to fit more into a certain form factor or the certain server or device. And so what we did is we created this uh ground up standard to be able to talk to, to be able to fit into some of these uh servers and new server usages, um, and so we created edsff in order to go solve that problem and it's kind of taken off on on on its own, because now it's not only being used for SSDs but we're also using it for, like, a CXL memory module. It's being looked at from a network form factor as well. So it has it's kind of taken off into its own usages.

Speaker 2:

And because of that you know what else are you seeing in general in the industry that is making use of this in the SSF world.

Speaker 6:

Yeah, so I mentioned the SSDs, cxl memory and the network devices. Really, some of the other things are what are we doing about AI, basically, what do you do about some of the processors for that? And so we've been looking at solutions for that. We've been looking at cabling to connect to some of those devices, whether it's copper or optical interconnects, and so our transceivers group has been working on updating specifications to not only hit, like the current speeds, but what are we doing in the future? For you know, 224, 448 plus, right, so a lot of these higher speeds.

Speaker 2:

I'm still holding out for ATM at the desktop. I'm old school. I remember working at the financial world and we actually had ATM cards going to trader desks. So it's been surprising where I guess that's for me as an outsider looking in. That's what's been fantastic to see the birth of these standards in SNEA and then the innovation that occurs through the industry partnerships, that occurs through the industry partnerships, and then sometimes what's old becomes new again and we see, because of changes in patterns of applications and infrastructure, none of these standards are there for a reason, because then we have a well-defined method and group of folks that contributed to that collective standard.

Speaker 2:

It's very, very difficult in our world where people can just pop up a startup, come up with a brand new idea and then go completely independent and proprietary and then the risk there is, then the ability to innovate is kind of capped. You can do a lot, but ultimately it's whatever the proverb would be is that we get there faster by going as a group than going alone. So it's been interesting to see and let's move on. We'll talk about, because I haven't got enough puns left in my world. Should I stay or should I go? Sta, anthony, what is STA and what does that come into play around the SNEA world?

Speaker 6:

So ST sta is the scuzzy trade association um, but basically their mission is to promote the understanding and use of uh, serial attached scuzzy um. And really what they're doing is they're trying to influence the evolution of scuzzy standards to meet future industry needs now, what do we do?

Speaker 2:

you know what are some are some of the meetups and things that we do in order to drive some of those conversations, because folks may not, if they're brand new to SNEA, they may not know kind of what type of stuff. I know we've got plug fests, we've got other things that we do. What are ways that people actually collaborate on these ideas?

Speaker 6:

I mean, I think the biggest one is probably the plug fest. Right, there's SAS plug fests that occur every so often that people can kind of come in into a room and do effectively a you know, does my device work with this, does my device work with that, you know, and kind of come in and go and flush some of those, you know, maybe flush problems out, maybe flush out some concerns that they've had and being able to go see that. And then of course there's also obviously joining stay.

Speaker 2:

And luckily it's super easy to do. That's one thing I found SNE is fantastic of being able to reach beyond your current column of people you talk to. It's easy to spread across the practices and pillars obviously a big area of work. So, fred Craig, what are SNEA's activities that are using NVMe and innovating and building around that?

Speaker 5:

Well, there are several activities that SNEA has partnered with NVMe.

Speaker 5:

An interesting one is the computational storage area, or having compute near the data, and NVMe has worked on the project at the protocol level to enable the interaction between the operating system, the host, the applications and the computation that lives in the storage device and the computation that lives in the storage device.

Speaker 5:

And SNEA has been working on application layer APIs so that applications can take advantage of those capabilities in a standard way.

Speaker 5:

So we have at the SNEA layer, the SNEA organization, we build the standards for the layer of the applications, for the device drivers, for all the communication within the host, the layer of the applications for the device drivers, for all the communication within the host, and then NVMe is working on the protocol to carry that information from the host over to the storage device where the operations can actually occur.

Speaker 5:

So it's a very nice kind of cooperative effort between SNEA and NVMe on how that having the compute near the data, the computational storage, work together between the two organizations. We've also developed within SNEA something called the NVMeOF drive specification, which is a common connector, a common way to put a bunch of NVMe OF drives together and have those connected up over Ethernet or something such as that, and that is a common specification that SNEA built but which operates with NVMe drives in that packaging. So you can imagine a shelf that has a bunch of either straight PCI NVMe drives that has an Ethernet front end to talk to a host, or maybe it's a bunch of Ethernet NVMe OF drives that are all in that shelf, that are all connected then to a host where SNEA has been the designer of that specification, packaging and connector, whereas it's the standard NVMe protocol that's used to communicate to that shelf of devices.

Speaker 2:

So it also, as I say, on the management side. Now then that becomes the. You're not even just at the protocol layer, but the management layer, which is pretty nifty. So I saw Swordfish pop up in the talk track. I'm like, oh, rochelle's ears will be burning in a great way. What's the connection with Swordfish and NVMe?

Speaker 5:

Well, swordfish is the storage part of the Redfish management operations that are for entire platforms. So the Swordfish part is the storage part and that is developed by the SNEA organization and so we interact with a number of the protocols. Nvme is, of course, the latest, but we also include management of SCSI and other storage protocols in that Swordfish management capability protocols in that Swordfish management capability. So we also have something called SDP that we're working on within SNEA. That's storage data placement.

Speaker 5:

Everyone probably remembers from the hard drive days when we talked about cylinder track and sector, where you could design applications that knew something about the geometry of the hardware and you could take advantage of that to build a faster access to that storage. Well, with solid-state drives we don't have cylinders track in sectors anymore, but we do still have fancy hardware components involved, and having some level of knowledge about that or being able to provide knowledge to the drive about the data itself are ways that additional performance can be eked out of those drives. So the SDP program is a way of looking at the zone storage that's provided by both SMR drives as well as the zone devices within SSDs. It can also apply to the individual die level within an SSD through something called. Oh help me, bill, what's the name of the? Is it FDP? The name within NVMe.

Speaker 3:

Yes, flexible data placement.

Speaker 5:

Yes, the flexible data placement, where the host can indicate to the device that there are certain sets of data that are related to each other and so they should be placed close together within the SSD architecture so that the management operations within the device can get some better performance by knowing that information. So the way to To do that is again through an application layer API, which is what SNEA is developing in its storage data placement specification, to be able to provide a way to interface to the storage level protocols, which are either the zoned devices or the FTP, the flexible data placement devices, the FTP, the Flexible Data Placement Devices. So, again, a cooperation between SNEA managing and operating the data and providing information and attributes about the data to the lower level protocols, which could be either be SCSI or NVMe.

Speaker 2:

One of the things that was fun about SDC. A couple years ago I remember bringing some folks who were, you know, I'd been working with in the industry, who were on the software side, in and introducing them to folks at SDC and it was neat to see where work like stuff that's happening with FTP, there were some people trying to tackle it on the software side not realizing there's opportunity where there's work that's being done on the hardware layer and it's. I guess we look over decades. Right, when everything came off at first of Flash, we worried about what's the life cycle of it, and then the more and more that we drive this innovation on increasing the ability, we can reduce wear, we can increase sustainability.

Speaker 2:

And for the folks that were around two years ago, when sustainability was everybody's, it was dripping off the tongues of every vendor marketer, we have always been working on sustainability and at the protocol and at the placement layer. That's where I think is really cool that people don't always see this stuff on the outsides, you know, in the consumer market. But this is the kind of work that allows us to really stretch the life of hardware and have a true impact on sustainability. And since I'm old school and I'm going to go back to my FC world. Let's talk, craig, on what has been done with around FC and what is going on with stuff that makes FC and the definition of some of those specs important in what SNEA is doing.

Speaker 4:

So SNEA and Fiber General have worked together for a long time, going back to some original management profiles that were done, and this was probably 15, 20 years ago. All the way up to today, the work is taking place around security. Snea and Fiber Channel have been working on a joint security white paper. It's actually an update of the white paper that's been around for a while and security protocols change based on best practices and so um, currently um, they've been working in our general community, have been working together to update the existing white paper, um, and and that the knowledge around that with the existing best practices, that that, that um, the security community is, and the good thing too is for folks that maybe don't even recognize all this stuff that's going practices that the security community is using.

Speaker 2:

And the good thing too is for folks that maybe don't even recognize all the stuff that's going on. We're going to make reference to some of the white papers and other work that's been published. If you go to the SNEA site, of course you can check through the document library. There's lots of information there. We'll include some links in the podcast notes as well and, of course, anybody that wants to reach out just always hit up the folks at SNEA and you'll find there's a lot of stuff that's been done and great reading and research that we can tap into before we think we've got a new idea going forward.

Speaker 2:

And then it goes beyond this because it's not just within SNEA but the partnerships. And that's why CXL was interesting, because of course CXL is fairly far along in some of the work that's been done. You know the specification has moved up in version numbers quite a bit since I've been watching it and since it's been around. So, craig Anthony, if you guys want to sort of pick and choose who wants to talk about what is the relationship with SNEA and CXL and what you see as kind of neat things that are coming ahead for us.

Speaker 4:

Well, you know it's still being explored for CXL, but some possible future tie-ins might be into the SDXI workgroup, which is a SNEA data mover. There might be some interesting things working with CXL and that, and so SDXI is a standardized data mover that allows you to have a hardware data mover that software, more than just a proprietary driver, knows how to use, and when you're talking about moving data in between pools of memory in CXL, that could be a good fit, and so that's just one of the areas that are possible fits for CXL. But it's a new technology so it's still open as to what exactly all the different applications could be.

Speaker 3:

It's still open as to what exactly?

Speaker 3:

all the different applications could be. So let me jump in there with one other thought on that as well. One of the things we are looking at in SNEA is we're looking at how do you merge three different things, how do you merge CXL, computational storage and SGXI storage and SGXI and you can take all three of these things and merge them together to provide an environment where you have different types of data movement, you have coherency and all three protocols working together. And we're looking at that currently within SNEA. But we're looking at taking two of our technical workgroups the computational storage workgroup and the STXI workgroup and working together on that and how we incorporate CXL with that as well.

Speaker 2:

Nice and Anthony, you had something to add as well.

Speaker 6:

Yeah, and we've had a little bit of collaboration there on terms of the CXL memory side. I mean, I mentioned the CXL memory module before. That was a collaboration between the SFF portion of CINEA and JADEC, jadec doing the memory side and then CXL being involved from the protocol transport side and then us from the physical hardware side, from the protocol transport side and then us from the physical hardware side. So it was a really good industry collaboration and a really good example of what we can do when we kind of all put our heads together to kind of go solve a problem.

Speaker 2:

Exactly, and this is again why I've been a longtime fan of SNEA. It's a bloody honor to be able to be the host of the podcast because I get to hang out with amazing folks, like all of you, and see as we go, to watch this evolution over time. And, as you said, even the organization evolves as we see working groups begin to sort of merge and collaborate and cross over. It can't be done in a vacuum and certainly can't be done in an ivory tower. So it's great to see I know we've got one more quick thing to hit because we've got to wrap, and I wish I could do this all day with you folks. But let's talk about the far future and very early work. What about UEC, craig? We mentioned it a bit before. What is the state today of the work with the UltraEthernet Consortium and what do you see as the next stages for folks to be able to watch and get involved?

Speaker 4:

So right now UEC and SNEA are working on an alliance agreement and UEC is in a development phase. Right now no specifications have been released so at this point I can't talk about specifics within UEC Coming up. In the next six months or so you'll start seeing some more public information on that. But some areas that I think that SNEA and UEC can really work together on SNEA just started a task force to look at AI applications for SNEA and UEC can really work together on. Snea just started a task force to look at AI applications for SNEA and UEC. I believe could be a big part of that, especially in some of the storage areas. But you know, without having everybody sign an NDA, unfortunately I can't talk more about that at this point.

Speaker 2:

Well and again, I think this is why this has been such a fantastic chat and I know it's been a bit of a speed run through things. But if I would go back to why we did this and why this is such a great discussion to have, we can see the breadth of coverage that's happening within SNEA, the amount of people that are collaborating and being there as we define what's the future. What will UEC look like? What will the collaboration look like, being able to get involved and to spend time with fine folks like yourselves. I tell people, just go jump in. Being a part of SNEA is a beautiful thing, it's not costly at all and it's also just nothing but positive as far as I've seen, and now that we're doing more and stuff back in person. We saw SDC in Austin, so we're seeing maybe even some more quarterly and more often and more regional events where we can collaborate. It's a great chance for people to get involved and jump in.

Speaker 2:

And again I want to thank everybody. You've all been great today, bill. Just one last thing. What's if you could collapse it into a couple of sentences why is all of this awesome and what is being an expert on data in the transport world mean to you. Oh, we got you muted there.

Speaker 3:

Thank you. I think the biggest thing is it's been said over and over we are collaborating with a large number of organizations that are all related to the transport of data and we bring them together, we work together with them, we bring them together, we work together with them and, as Anthony mentioned, it's even a matter of kind of being the centerpiece that brings other organizations together to enhance the overall industry in terms of how we transport data and we will continue moving forward with that as we move into the next generation in the UEC.

Speaker 2:

Amazing, yeah, and it's been an inspiration to watch. Thank you all for taking the time to share today. Again, for folks that are brand new to us, make sure you do go head on over to the SNEA website, get yourself signed up, join the community Watch. We've got the Experts on Data podcast is both in audio format. It's also on the YouTube channel, so if you're listening, pop on over and you can see these fantastic, beautiful faces directly. And if you want to see them in person, we've got lots of events that are coming up. Head to the SNEA page. Check out what's ahead and watch for more as we talk about all of the different practice areas. Thank you all, folks.

Speaker 1:

My name is Eric Wright. Again, thank you for listening to the SNEA Experts on Data podcast. Thank you for listening. For additional information on the material presented in this podcast. Be sure and check out our educational library at sniaorg slash library.