Host Intro

Welcome back to the LTO Show — the premier podcast for leaders in the tape storage community, where we go deep on the technology, the trends, and the people shaping enterprise tape and long-term data archiving.

I’m very excited to feature the innovative Spectra OSW-2400 Optical SAS Switch on the LTO Show today. I remember when I first read the initial product announcement a little over a year ago — it was one of those moments where you stop and go, “Wow.” I had no idea this was even a thing.

And it made me scratch my head for a while. I was genuinely amazed to learn that Spectra had engineered a replacement for the conventional storage SAN architecture — something built around SAS connectivity, which I’d always considered a peer-to-peer interconnect technology. It really stretched my brain cells to think of it as a storage fabric. And it was the kind of stretch that felt good.

It’s like hearing a new song you really like — it doesn’t happen too often, so when it does, you remember the moment. OK, I’m probably getting a little too emotional over a tech product in the storage industry. But some of you might know exactly what I’m talking about.

Anyway, since I started the LTO Show, I’ve wanted to do a deeper dive into the use cases and the technology behind this development. And I’ve really enjoyed learning more from today’s guest — Chris Bukowski, the Spectra Logic product manager — about how it works, and how it provides many of the same configuration and management capabilities that a Fibre Channel switch does on a SAN.

We’re talking fabric zoning to isolate servers and devices for security and traffic management — just like you’d do on an FC SAN. Multi-pathing and dual-port failover for redundancy. Integration with network monitoring platforms like Splunk. And up to 40 tape drives connected through a single 1U switch, with 1.08 terabits per second of aggregate bandwidth across the fabric.

What makes this especially interesting is how Spectra gets the distance. Rather than converting SAS to Fibre Channel — which adds protocol overhead and cost — they’ve kept the native SAS protocol intact and moved the signal over active optical cable at the physical layer. That gives you up to 100 meters of reach inside a data center, with none of the latency penalty you’d take from a protocol bridge. And the switch runs SAS-4 at 22.5 gigabits per second per lane, which means it’s already forward-proofed for the next generation of LTO drives — well ahead of where LTO-10 sits today.

It really is a thoughtful piece of engineering. And it raises some genuinely interesting architectural possibilities for how tape infrastructure gets designed going forward — some of which I think even Spectra hasn’t fully marketed yet.

My guest today is Chris Bukowski, Senior Manager of Product Management at Spectra Logic, where he’s been leading the development and go-to-market strategy for the OSW-2400.

Chris, welcome to the LTO Show.

Pete Paisley: Chris Bukowski, Senior Manager of Product Management Spectralogic. Thanks for joining us on the LTO Show.

Chris Bukowski: Thank you. Thank you. It’s great to be here.

Pete Paisley: I’ve been looking forward to this session for a long time, Chris. I was fascinated when I first heard about your SAS switch hitting the market a year or so ago. So excited to have you on the show and dig a little deeper into it. And I always like to start every interview with kind of the origin story. How creative is that? So if you could take a minute, walk us back to the beginning. What was the conversation inside Spectra that led to the OSW2400, your SAS switch? Was a specific customer hitting a wall, or did your engineering team see the gap before customers asked?

Chris Bukowski: You know, it was about saving customers money, right? So we have several different libraries, several different tape libraries that Spectre sells, and from very large to exabytes worth of data down to smaller rack-mounted libraries and even smaller than that today. But we found that that that a lot of customers with the very large libraries were using fiber channel networks, fiber channel SAN, and noticing that fiber channel SAN were almost to the point where the only thing that you were using a fiber channel SAN for anymore was a tape library. And even disk subsystems and things like that were done on the Ethernet network and not so much on a fiber channel SAN And so we were thinking, those SAN can become really expensive and what’s an opportunity to save our customers some money and give them the same functionality that they would see with a fiber channel network and we came up with this idea for a SAS switch, an optical SAS switch. So three of the things that we get from that SAS switch are just a lot of cost savings and I can go into that at a later point where the real cost savings are. so much easier to use than a fiber channel SAN. And now with a SAS fabric, you can actually have the same kind of distances within a data center that you can with a fiber channel network, where in the old days with SAS, you’re limited to three or four meters on a connection. And so it just worked out to be a great advantage for our customers that didn’t necessarily want to buy. a new fiber channel network when they didn’t already have one, right, that they didn’t want to link the library into. And so it’s a great cost savings for our customers.

Pete Paisley: Right. Yeah. And really there is a fiber channel tax today because, um, fiber channel tape drives cost depending on the latest price increase from IBM 1200, or even much more than their SAS equivalents. Right. And fiber channel switch ports are considerably more expensive, I believe today than, than a SAS switch in your configuration would be. So

Chris Bukowski: you

Pete Paisley: You help people with that tax, right?

Chris Bukowski: Yeah, absolutely. I’ve done a detailed analysis and we do this for a lot of our customers where now we’re… You’re seeing where they’re asking us for quotes, hey, what if we did this library in all fiber channel and what if we did this library in all SAS? And one of the things we’re finding is the SAS tape drives themselves. So an LTO 10 to LTO 10 SAS to fiber channel is approximately $1,000 cheaper for the SAS counterpart. And in the fabric itself, a fiber channel fabric or network compared to a SAS fabric, can save, it’s less than $50 the cost per port. Sometimes we can go down, depending on the configuration, we can go down as much as 70 % less cost than a fiber channel network. And that leaves customers with a lot of great options. So like the example that I mentioned before, where customers were asking for a fiber channel configuration and a SAS configuration, we can look at both of those and save them tens of thousands of dollars on a big configuration where they can look at they can say we can save money on our overall installation, or we can buy more tape drives. So when it was, say, maybe a 15 tape drive library with a lot of media that goes in there, and then if they do it in SAS, well, maybe we can do 20 tape drives for the same cost. And that helps their whole system when it comes to throughput and their data movers to the tape library. It’s worked out to be a great solution.

Pete Paisley: Yeah. And, I had a note here to ask about, you know, you launched this a little over a year ago, which kind of coincided with the push to repatriate cloud data, you know, move stuff back on prem, particularly cold archive where some of the costs people were seeing on cloud storage, particularly around egress costs were pretty punishing. And I don’t know that you’re announcement at the time was related to this, but do you think the market trends around cloud repatriation are, know, does it help that they have solutions like this SAS switch that help reduce the costs of implementing tape on-prem?

Chris Bukowski: It does. mean, in the storage world today. It’s all about costs. mean, the costs have gone up so high when it comes to flash and just spinning disk and even SATA drives. And the real cost advantages for meeting cold, warm, cold data, kind of long-term archival storage is the best option is tape. And so we are seeing a lot of attention for that. of what happened with cloud storage is as people realize, hey, I’ve got a lot of stuff in cloud storage and I don’t want to access it all the time or I don’t need to access it all the time, that’s when AWS comes up with this Glacier model. And really behind the Glacier model is tape libraries back there where they charge more for access and less for storage. They have that whole span of how they price things on one end.

Pete Paisley: Yes.

Chris Bukowski: It’s more for the storage and less for the access. And on the glacier side, it’s less for the access or less for the storage. But every time you access it, turns out to be very expensive. So if you’re putting things into the cloud for long-term storage or cold storage that maybe you feel like I’m going to get to maybe 2 or 3 % of this every year is all I’m going to access. And you put that in your budget. And all of a sudden, you decide you need to access it 7 or 8 % of the time or even 10 % of the time, your costs, all of a sudden you get extra multi-million dollar bill from AWS that, hey, that’s not what I was expecting here. I’m just putting this data up here. I’m going to access it once or two times in its whole life. So we are seeing a lot of repatriation. We are seeing customers that are saying, hey, how about if I just put a library?

Pete Paisley: Yeah.

Chris Bukowski: site in my own data center and can I save and say and you’re exactly right you save on egress fees you save on access costs you save on on just data migration costs that you might have and and you can have access to anything in a tape library you know if it’s if it’s if it’s actually you know out of the cartridge out of the tape drive and then a cartridge that’s in the library that you can have access to the data in a couple of minutes which is which is usually quick enough when it comes to cold data

Pete Paisley: Yeah, sure. Now, I bet your ideal customer for the OSW 2400 SAS switch is all your customers, but let’s target in a little bit on that ideal customer profile. Is it primarily a retrofit for existing customers who are running a fiber channel SAN, or are you seeing more Greenfield deployments where people are building tape infrastructure for the first time?

Chris Bukowski: Well, a couple of different ways. So one is we have, there are customers out there that are. that are trying to get off of their fiber channel SAN. So they’ve moved away from it for disk storage and secondary disk storage for SAN. And they might still have some that they’re using for their library. And they’d love to be able to get away from fiber channel altogether. The SAS switch gives them that ability to put in a SAS fabric that can do basically the same thing that the fiber channel SAN can do. up to a certain distance. And then there are other customers that don’t have fiber channel. And when it comes to a big library, their only option when it came to distance was to put in a fiber channel SAN. And people kind of try to stay away from that these days. don’t want to put in a fiber channel SAN if they don’t have one already. And so now we have this option to put in the SAS switch or multiple switches and create a whole SAS fabric that

Pete Paisley: Yeah.

Chris Bukowski: essentially does what a fiber channel SAN can do at half the cost.

Pete Paisley: Chris, before we get to the switch itself, help our listeners understand why tape is so unforgiving about data flow. What actually happens mechanically when the data stream coming in isn’t steady?

Chris Bukowski: anybody who has a tape library has heard of tape libraries that will do shoe shining or back hitching back and forth and that comes from the data stream not staying constant with the tape drive itself. One of the things that we’ve designed into the SAS switch itself is the capacity for the SAS switch is enough that any data stream will be able to keep current and keep constant to the tape drive itself, not alone the tape library, and to many drives itself. So the switch itself has 48 lanes in a single switch. It’s a single 1U switch. And of those, you can connect. one or two ports back to the host and you can actually keep as many as 40, 36, 40 tape drives streaming through the SAS switch without any hitching or any shoe shining.

Pete Paisley: That seems fantastic. So on that note too, I noticed that your switch is SAS4, right? And tape drives today are SAS3. So on the surface, seems like a speed mismatch. how does, you alluded to it just then, but how does the fabric handle differences between host side and drive side speeds while maintaining this steady streaming rate for tape?

Chris Bukowski: Well, there’s a couple of things built into the switch that we use. But one of the things I should mention is that in the tape drive world, we pretty much stay with SAS 3. And one of the reasons for that is in the SAS 4 chips, There was a piece, the piece that was designed out of the SAS4 chips that is key to a tape drive installation and it’s called TLR. And TLR is only built into SAS3. So for SAS, the SAS4 capabilities for the switch are for, we have in there for future deployments in case TLR is adopted into SAS4 at some point in the future. It may or may not, mean, SAS3 speeds are really good enough for tape drives today. And it’s just a matter of

Pete Paisley: Yeah.

Chris Bukowski: how many that you can put on the switch itself. And then for SAS4, we’d also use those when we connect the switch up to, a JBOD or a DERP. some SAS disk library, right? So you can connect to it with SAS4. Now most of the drives are still SAS3 also, but in order to accommodate that, there’s a couple of different things that are built into the switch. One is EDFB. EDFB is the ability to help, you know, it’s a scheduling mechanism, but it helps you to make sure that the load is balanced across the switch because you’re

Pete Paisley: Right.

Chris Bukowski: The switch itself internally works like you have multiple lanes coming in and then more lanes going out to the devices themselves. And the EDFB will help you balance that load across the lanes that you’re connected to on the devices, whether they be tape drives or disk drives. Today we have, we use SAS 3 to the tape drives and all the tape drives today since LTO 8 have been SAS 3. Prior to LTO 8, they were SAS 2. So if you have, say, a SAS 3 HBA connected to SAS 3 tape drives, you don’t have anything to worry about. You don’t have any… contention between different speed levels and typically those lanes will go to the lowest speed level of the device or the initiator. One of the other things we have built in is what we call data bolt. And data bolt is a buffering mechanism. if you do happen to have a SAS4 host and you’re connected to all SAS3 or SAS2 devices, the switch can actually buffer that. throughput, so you can push it. The data mover can push the data as fast as it can into the switch, and the switch can buffer that data out to the different speeds or the different devices that are, say, at a lower SAS level, SAS 3. So you don’t have to have the whole lane connection from the initiator to the target go all the way down to SAS 2 or SAS 3. But in the tape drive world, and what we’re seeing today is most of our installations are LTO 8, 9, 10, the newest drive. And SAS 3 HBAs are the cream of the crop for tape.

Pete Paisley: Yeah.

Chris Bukowski: And then we don’t have any of that contention. So you can utilize a single connection back to a SAS 3 host, a single port connection that has four lanes on it. And you can accommodate full throughput at up to 12 tape drives with a single connection back to the host, where in the old days, before the SAS switch, you had to have a single connection to the host. And it would be in a copper breakout cable. And that breakout cable would go to four different tape drives and use a single lane for each of the tape drives. Now we can do that single connection to the host with up to 12 tape drives instead of just four.

Pete Paisley: So you walked us a little bit through how EDFB works. but essentially it, it’s helping this produce a smooth traffic flow into the tape drive, a more consistent stream. I find it interesting, too, that you mentioned TLR support on SAS 3, because that’s a critical element for LTFS environments, right? Yeah. Yeah. So isn’t there a known challenge with EDFB where the error tracking burden across many ports can actually eat into the performance benefit? I know you mentioned how a single host can serve multiple tape drives smoothly and efficiently.

Chris Bukowski: It is.

Pete Paisley: on your SAS switch. But are there some trade-offs for EDFB implementations around the overhead that can offset some of the gain?

Chris Bukowski: Well, not in the tape drive world. In the tape drive world, when we’re going SAS 3 to SAS 3, there’s really no need for it. So there’s no need to necessarily do that balancing because you’ll typically have the.

Pete Paisley: Okay.

Chris Bukowski: the speed of a tape drive is about 400 megabytes per second, and the speed of a particular lane is 1.2 gigabytes per second, right? So you can actually have multiple tape drives on the same lane if you wanted to. where it really comes into play is more of if you’re doing a SAS4 connection or a SAS3 connection to multiple devices that are sharing the load and you don’t necessarily have the capacity for all the disk devices that are on the other end. And so EDFB stands for Earliest Deadline First with Batching. So it’s really just a fancy name to help do the load balancing and the scheduling when you’re connected to a few lanes to many lanes in the SAS fabric.

Pete Paisley: Interesting. Early deadline, first batching. Is that what I heard?

Chris Bukowski: early deadline first with batching, And

Pete Paisley: first with matching. Okay. Great trivia question. that’s a great question to ask at a bar. You’ll get them every time. I love it.

Chris Bukowski: Yeah, right. Yeah, see? And then that, combination with DataBolt, have, DataBolt allows you to buffer, right, to do the buffering. And those two elements really allow you to keep, it helps you get a best case streaming with your DataMover and all your data to your devices.

Pete Paisley: Yeah. Yeah, now I’ve kind of been dying to ask, because in my research for this, it seemed pretty clear to me that what’s been done that’s really cool here is it’s a SAS switch, SAS fabric, but the physical layer is running over multi-mode fiber or single-mode fiber or something. So you can get that distance. Is that basically what’s going on here? After it leaves a switch, it’s converting. to optical for the transmission to the other end and it converts back.

Chris Bukowski: Yeah, I mean, that’s all done within the active optical cables and the connectors itself. So it’s multi-mode fiber, right? It’s not single mode. It’s multi-mode fiber. And it has unique connectors on the end that…

Pete Paisley: Yeah. Okay.

Chris Bukowski: make it so it’s all within the cabling itself. Now all you have to know in the HBA or in the switch itself is that it has enough power to the ports to be able to handle that active optical cable or that active optical signaling. And all the HPAs that we use here at Spector are that way. But there are some out there that don’t have the power necessarily to drive an active optical cable. Now one of the things too on tape drives is until LTO-10, I think through LTO-9, LTO tape drives couldn’t handle an active optical cable. They don’t have that power built into the connector to be able to handle that. But with LTO-10… With LTO-10, I’m pretty sure that that is there now. So if you wanted to, you could go ahead and connect an active optical cable from the switch to the tape drive itself. But most people don’t do that because typically what happens is the switch itself will be right next to the library or built into the library itself. And then you have copper cables that are connected to from the ports on the switch that are connected to the tape drives themselves. And the active optical cable is from the switch on the other side of the data center where your servers are where your data movers are.

Pete Paisley: Yep, yep. And on that note, I was thinking you could have taken a different approach. You could have converted SAS to fiber channel and back. You could have just gone with a straight fiber channel implementation over regular fiber channel cabling. And it could have given you some longer distances than you support with the unit today, right? What did you gain by staying at the native SAS protocol layer? And where does that approach make the most sense versus longer distance alternatives like Fiber Channel?

Chris Bukowski: there’s always overhead involved in changing protocols. So if you’re changing from fiber channel to SAS or SAS to ethernet or whatever, there’s always some performance overhead in doing those transfers. And we found that we get our best performance and our best.

Pete Paisley: Mm-hmm.

Chris Bukowski: all the way back to what you said about the shoe shining and all that with the tape drives and back hitching. In a complete SAS network or a complete fiber network, you get much less overhead for that protocol transfer or that protocol change. And so that’s why we’ve kind of seen better performance when we create a full SAS fabric to a SAS tape drive than if we put in a bridge that’s a connected SAS tape

Pete Paisley: Right.

Chris Bukowski: drives to a fiber channel network

Pete Paisley: So the 100 meter range of your SAS switch, is it working for most real world use cases where the customer wants the library to be in a separate location, for instance? It hasn’t been a constraint for you.

Chris Bukowski: It is. fact, we have a… No, no, no. So there’s a couple of pieces to that is, you know, we have different optical cables at all different distances, 30, 50, know, 100 meters. And that’s for a single connection between an HBA and a switch or between two switches. So you can create a SAS fabric with cascading switches up to, and we’ve tested up to a kilometer. So you can actually string these switches together like repeaters in a SAS network.

Pete Paisley: There you go.

Chris Bukowski: and have 100 meter cables between up to 10 switches and be able to, and that gives you great flexibility for your fabric all the way across your data center. If you had to move SAS around and your servers are in one rack over here, but your tape library is all the way on the other side, but maybe you need to connect to some JBODs here in the middle, you can create a fabric of multiple SAS switches all connected through to the same host,

Pete Paisley: Sure.

Chris Bukowski: multiple hosts, can have failover, multi-pathing through the switches to each of the tape drives, because the tape drives do have dual ports. They have dual SAS ports, dual fiber channel ports, right? So you can connect that way. But it’s a great way to get extra distance beyond the 100 meters themselves. And that’s been huge for our customers is just the flexibility to be able to have your servers on the other side of the data center. from where your tape library is, or maybe even in another room, because you have maybe a special environment room for your high performance servers than you do for your tape libraries.

Pete Paisley: Chris, here’s a use case I just referred to. I’ve been thinking about that. I haven’t seen you really market this strongly yet, but modern tape libraries at exabyte scale really benefit from being in a dedicated controlled environment. LTO9, LTO10 drives are more sensitive than ever to temperature, humidity, particulate contamination, et cetera. And a HEPA-filtered positive pressure precise humidity control environment can be hugely beneficial to the performance and the longevity of tape backup systems. Now, IBM recently announced a diamond back library climate control model where they essentially put an air conditioner on the back of the library wherever it goes. I kind of think of that as the spacesuit model, right? But the OSW2400 enables something architecturally different. You can put the library in a purpose-built controlled room, run your SAS connections to it off your switch or switches to wherever your servers live up to 100 meters away. And I kind of think of that as the boy in the bubble model. And I think it seems more cost effective. Is that a deployment model you’ve seen any customers explore? built around your Switch.

Chris Bukowski: It is. mean, you know, in the, the We do a great job of handling the environmentals in our libraries and we try to minimize the effect of any sort of variables in the environment that way. But one of the things that the switch allows now is that if you do decide to put your exabyte size or your large library in a specific room or in a specific area that has more air conditioning or less, the side note is that sometimes server, high performance servers don’t live in the same environment that your tape library does, right? And so what you can do is you can put the switch there and you can move those servers, those high performance servers off to another part of the data center or even another room. And we have seen customers that are looking at that. In fact, we have one situation or one proof of concept that’s going on right now where the the customer has servers in multiple different rooms in their building, and the library is in a separate room. And we’ve got 100 meter cables going from different servers all over their building into a SAS switch that’s going to connect to their library. And the library is in a different environmental room than the servers are. And in a way, it’s what you’re talking about. So it’s a way to isolate that without having to have, you know, SAS environment today, the servers and your data movers have to be right next to your library. And we don’t have to do that anymore.

Pete Paisley: Yeah, and I also noticed, I think Black Pearl, the most recent version, added support for the OSW 2400. And since Black Pearl manages both disk and tape in your tech stack, so it’s kind of connecting the fabric then in my mind. Are you seeing customers extend this SAS fabric beyond tape into these adjacent storage tiers where maybe a A warmer cold storage tier on disk might be SAS attached using this switch sitting in front of the library. Or is it has it extended to adding disk into the mix for the connectivity for this device?

Chris Bukowski: It is. with the Black Pearl, have, you the Black Pearl is a object storage platform. object gateway as we call it. And what it is, it’s a great object storage platform where you can go petabytes worth of disk or flash on the Black Pearl itself with a gateway to a tape library. object-based tape. And so now in your object storage platform or your object environment that you have, you can have quick access to disk or flash within the Black Pearl itself or with, integrate with other systems to other object storage systems. And you can also go to, you can have your cold storage or you can set up policies where as something gets to a certain age where you were before you would roll it off to Glacier and AWS or something like that, you can roll it off to your tape library and you can get it back as quick as a couple of minutes, right? And on the disk, you can get it back in milliseconds, right? So it’s helpful. helped us to expand the flexibility of that whole environment where you can have your Black Pearl and your object storage gateway in one part of your data center and be able to have the distance to get to other disk subsystems, other SAS JBODs, or to a SAS tape library.

Pete Paisley: Yeah. Well, and in today’s environment where spinning disc SSD is really hard to get. tape will probably become a warmer tier of storage. And to do that, it has to be enabled with a caching layer of disks, such as Black Pearl Accomplishes. And then it makes sense that it would be part of that same SAS fabric for the connectivity. I also want to ask, do you have any early customer deployments or real-world installations that you find of interest that

Chris Bukowski: It is.

Pete Paisley: that surprised you about how they used it, anything that stands out.

Chris Bukowski: You know, I think the… The biggest surprise that we’re getting out of this is how receptive people are to going with a SAS fabric. Because it’s a new thing, right? mean, a SAS world has been direct connections and point to point connections to either JBODs or, again, to tape on the smaller scale. But one of the things we’re seeing is great receptiveness to. to reducing the cost of having to do fiber channel. And that’s been pretty common. In particular, I think the thing that surprised me too is as customers hear about this, it’s like, all right, well, give me a quote compared a fiber channel subsystem, SAN with my fiber channel library and a SAS library. And when they really look at the benefits of the SAS fabric can do everything that the fiber channel fabric can, except for going miles. And when it comes to within the data center, they’re looking at that. And then they’re saying, you know what? With my budget, I was budgeting for this fiber channel setup. And now I can do a SAS fabric setup. And I can actually buy more tape drives, which helps my data movers, which helps my data flow, which helps my whole workload.

Pete Paisley: Right.

Chris Bukowski: and I can get more done in less time for less money.

Pete Paisley: Chris, SAS has always had a reputation as more of a kind of server-based technology. Now that you’ve broken the distance barrier with your new SAS switch, how does that change the conversation? Could a customer under 100 nodes actually build a low-cost, high-performance SAS fabric for their entire data protection tier, not just the tape? Kind of like I was mentioning before, do you see this potentially expanding across more of the disk storage tiers in some data centers?

Chris Bukowski: At some point I think it will expand even more as people see the benefits of how you can create a SAS fabric and that it can do almost everything that a fiber channel fabric can do. Initially, we’re rolling this out strictly for Spectra tape libraries. But it will definitely expand into more of a full data center fabric kind of role. And one of the things that we’ve built into the SAS switch itself too is you can zone in the switch just like you can with a fiber channel SAN. So you can do whatever you need to when it comes to isolating particular servers particular tape drives or particular libraries. And there’s that data isolation is a great security feature to be able to keep different servers away from even seeing the other servers on the network itself in the fabric. And we also have the ability within the switch to integrate with networking applications like Splunk or things like that that can monitor your network and see where there’s the traffic and you can get updates on your SAS fabric into those kinds of applications.

Pete Paisley: I also wanted to ask because you guys sell Atto’s bridge product, their ethernet bridges. Do you see the SAS switch replacing some of your dependency on these third party ethernet switches?

Chris Bukowski: Maybe not necessarily. With the growth of the SAS end of the business, as more companies are moving away from Fiber Channel, we see it just as a way to make sure that we’re positioned to be able to cover all of that kind of, that new business aspect of it. So we have. There’s the SAS to fiber channel bridge that ATTO has, and there’s also the Ethernet to SAS bridge that they have. And we use those in several of our installations. In fact, we have one where we’ve got an Ethernet connection from over 200 kilometers away to our library through that same bridge, which we used to call the swarm. And so we still sell those when it’s right for the customer. But as more customers are looking at newer fabrics as opposed to adapting to older fabrics or existing fabrics, the SAS switch comes into play more and more often.

Pete Paisley: And I was also wondering if you go into an environment where a customer has a legacy fiber channel SAN, it’s fully paid for and amortized switches, HBAs, people who know it on staff, are they open to talk about a new SAS fabric for the storage area network?

Chris Bukowski: They are. In fact, we have some customers looking at they have a full fiber channel library. And they want to keep what they have on their fiber channel SAN. And they want to expand the library with SAS. And they’re going to use that so it takes a single HBA for up to 12 more SAS tape drives.

Pete Paisley: Mm.

Chris Bukowski: it connected through the switch and with all the distance advantages that you have. And one of the things you’ll see too is, like you mentioned before, the expense per port for a fiber channel SAN. Sometimes that can be as much as $2,000 per port when you talk about the cables and the HBAs and all the connections that you have to make. And we’re seeing that the per port cost on the SAS switch can be less than half that cost if not if not 75, 25 % of that cost, right? 75 % discount. And so when we show that to the customers, they’re like, well, this is all working. This fiber channel piece is all working. I’ll keep that. But when it comes to expanding, I’m going to look at as opposed to buying another fiber channel switch. And if they have dual porting, they’ve got to buy two fiber channel switches. And as opposed to doing that and spending $2,000 a port, approximately, they can do it with SAS for less than half the price. so you mentioned that before, but that’s one of the things that surprises me too is we’ve got a lot of customers, several at the moment, looking at expanding their fiber channel libraries with SAS and not necessarily converting the whole library to SAS, but keeping their investment in fiber channel, but their future growth plans go with SAS.

Pete Paisley: So LTO10 drives ship today with 12 gig SAS 3 interface. We kind of covered that earlier in the conversation. Plenty of bandwidth for 400 megabytes per second native transfer rate. But the OSW2400 switch runs SAS 4 at 22 and 1 half gigabytes per second, already a full generation ahead of the drives it serves today. I’m sure Spectre was thinking. We’re going to build lots of headroom into the switch. So it’ll support future generations, right? Can you expand on that a little bit for us? How you looked at the overall bandwidth, aggregate bandwidth of the switch in terms of the LTO ecosystem.

Chris Bukowski: Sure. So we went with SAS4 with because. The chip itself can handle SAS 3 and SAS 2 also at the same thing. So it makes us future proof. So we can actually handle the growth in the future of LTO drives that might go to SAS 4. Again, there’s the TLR question for that. But if the drives go to higher speeds, they might go to SAS, even stay at SAS 3 or SAS 4 and go to 800 megabytes per second when it comes to compressed speeds.

Pete Paisley: Yeah.

Chris Bukowski: the tape drives themselves. And then as disk drives go to SAS4, we have all that capacity built into the switch. So a customer that buys the switch today can actually knows that all that is built in and all they would have to do is maybe go to a SAS4 HBA and handle any SAS4 devices on the end. So future-proofed in that sense. But when it comes to the LTO roadmap itself, you’ve got LTO 11 that’ll probably be out

Pete Paisley: Yeah.

Chris Bukowski: in a couple of years and it might be faster. If it does have the ability to run with TLR in SAS4, we’re already positioned to be able to handle all of that and we’ll be in good shape to accommodate any of those bigger, faster drives.

Pete Paisley: Yes. Yeah, and that’s an interesting aspect to me because I’ve kind of been quizzing IBM every opportunity I get about LTO 11. We’ve talked about it on the LTO show before. Western Digital Pub, you know, they they published a paper about the 64 channel head design. IBM kind of acknowledged that that might be next. And since they’ve been on the 32 channel head for a few generations now, it continues to make sense. The question, though, is Will they double the bandwidth to 800 megabytes per second, or will they stage it? And now I’m kind of hearing that they’ll probably stage it. They won’t give it all away at first, which we understand. But yeah, it’s a great story for the SAS switch that it has more than enough capacity in terms of aggregate bandwidth and individual channel bandwidth to support that kind of new throughput. And personally, I’m excited about what a significantly faster drive could do in today’s data center environment with some of the performance requirements we have due to the massive amounts of storage people are needing to support. And again, back to my earlier proposition that tape will move up from a true strictly a cold tier to maybe a warmer tier of storage enabled by these faster speeds that you support with your new SAS switch. So, hey, I wanted to walk through some quick math for our listeners. As an example, future looking. You got 40 of these LTO 11 drives and let’s say they’re running 800 megabytes per second. Let’s say they just give it all up at first, right? That’s about 32 gigabytes per second of aggregate throughput. The OSW 2400 based on my calculations delivers 108 gigabytes per second.

Chris Bukowski: Okay.

Pete Paisley: across the switches fabric, right? Which is about three X the head, is that about right? Okay, cool. Yeah, it’s asked for, assuming they worked that out, which is about three times the headroom needed for those full 40 drive LTO 11 deployment at double the speed of drives today. So I guess we already covered it, but does that level of…

Chris Bukowski: at SAS4. Yes, at SAS4, yes.

Pete Paisley: of Headroom align with how you think about scaling for future generations with this product.

Chris Bukowski: Yeah, absolutely. mean, it’s, you know, if you think about… If you think about the SAS 3 world today, the drives are 400 megabytes a second and you’ve got 1.2 gigabytes per second per lane and we’ve got 48 lanes in there. So a single lane by itself could handle four tape drives at full speed, right? So if they double the tape drive speed to 800 megabytes a second, a single lane at SAS 3 can still handle that no problem and you can still do 48 or 40 drives with the

Pete Paisley: So yeah.

Chris Bukowski: with a couple of connections back to the host itself with no performance degradation, full throughput, and not have an issue there at all. So then if you make the jump to say the 800 megabytes a second is with a SAS4 and we solve the TLR issue itself, the SAS switch is completely capable of handling all of that throughput and still have know, as many as 40 tape drives connected at a single time.

Pete Paisley: Yeah. I wanted to, on that note, to talk a little bit about a hyperscaler question, right? So a lot of people outside of our space don’t fully appreciate how much tape is probably running inside a hyperscale environment. Meta’s storage team has publicly talked about exabyte scale tape archives. Can you give us a sense of what that world looks like? What role tape actually plays in these hyperscale infrastructure environments at that scale?

Chris Bukowski: Yeah. So. So one of the data points we like to use is, especially in these hyperscale and AI environments, once the data is processed, there’s as much as 60 some percent of the data becomes cold. So you use it once and then you process it and it’s perfect for sitting on cold storage or cold to warm kind of storage for access. We have environments where we have systems in exabytes of data that’s already on tape. With all of that in a hyperscale environment, all of that has access within a couple of minutes to be able to get to that cold data where you don’t have to put it on high cost. NVMe, high-cost flash, a high-cost spinning disk that also can’t handle that data for longer periods of time. mean, a tape cartridge can handle 25, 30 years worth of storage sitting in a tape library. And to have that access to that data with that quality of durability is really second to none when it comes to a tape library. And so we’ll see as the massive amounts of data get created, that just means that the amount of cold data mushrooms into even more need for something that’s onsite, ready access, maybe not immediate access, but kind of an archival cold storage that is economical and can last for decades.

Pete Paisley: Yeah.

Pete Paisley: We alluded to this before as flash and SSD costs remain high and increasing and disk supply tightens, there’s a real argument that tape is beginning to move up the storage hierarchy a little bit. Some people view it less as strictly cold storage and maybe more towards warmer use cases with more frequent access. If tape continues to warm up, what does that mean for connectivity requirements? It does seem to help the use case for the OSW2400SAS switch.

Chris Bukowski: it’s been traditionally cold storage, as it moves up to more of a, or. you as you’ve said, it tends to move into the warmer areas of the storage hierarchy. It’s because the cost of disk and flash have gotten so expensive. And so people are looking for lower cost alternatives where they can potentially wait a little bit for the data to come out. say you do a call to a tape library and it has to grab the data off of a cartridge that’s not in the drive, you might wait a couple of minutes to get that data. If you do that with, say, AWS Glacier, at particular classes of storage, you might have to wait 48 hours for that data. What you’re seeing is more and more applications understanding that that cold data is at a S3 Glacier level. That’s what we have built in with our Black Pearl, is an object gateway to to the tape library. And so you can set up your policies. So data stays on disk for 30 days and then it can migrate off to tape after 30 days and you can get very complex or detailed with that kind of storage. And applications need to know that, hey, I’m going to access something that might be on a tape library. And by the way, I might have to wait for three or four minutes to actually get the data back to be able to go do their media manipulation or whatever they do with the data. a lot of applications trying to adapt to that new kind of. paradigm So if it doesn’t have to be immediate high performance, low latency kind of data access for your application, if it can wait for a couple of minutes to get something off of a tape cartridge, that’s how the tape libraries would be moving up into a warm level of the storage hierarchy.

Pete Paisley: Yeah. Well, I congratulate Spectralogic for introducing the SAS switch. I think it’s a little beyond evolutionary. It is somewhat revolutionary not to overuse an overused term. I think connectivity is becoming a more important factor in how tape is deployed in modern environments. And this certainly provides. a lot of options for design purposes.

Chris Bukowski: We’re seeing great interest in conjunction with our libraries. It’s a great time in the storage business. Other than the price increases, it’s really a dynamic time in the storage business.

Pete Paisley: It is. Yeah, But I am interested to watch and see if your SaaS switch concept expands over time as kind of a standalone product family. Chris, I’m not aware of any other company that’s introduced a similar product even for other environments, right?

Chris Bukowski: We’re the only ones that we know of, It’s exclusive to Spectra tape libraries right now, and it’s exclusive to the market itself

Pete Paisley: Yeah, I’m really interested to see where it goes. So thank you so much for coming in today and talking about the Spectrologic SAS switch. And we’ll look forward to talking to you some more in the future about it as things progress.

Chris Bukowski: Thank you. Thank you. This has been great. I’d love to do it again.

Closing Thoughts

Chris, this has been a fantastic conversation — and I think it’s one of those episodes that listeners in the tape storage world are going to come back to.

What struck me most is how far the thinking here goes beyond a simple hardware product. The OSW-2400 isn’t just a cheaper alternative to a Fibre Channel switch — it’s a rethinking of how tape infrastructure is architected. The ability to zone the fabric, support multi-pathing and dual-port failover, integrate with monitoring tools like Splunk, cascade switches for extended reach, and run the whole thing over native SAS protocol without conversion overhead — that’s a serious fabric story.

And with SAS-4 already built in, the switch is positioned to handle the next generation of LTO drives — whatever speed LTO-11 lands at — with headroom to spare.

For our listeners who want to explore further, product documentation and specs are at spectralogic.com — we’ll have a direct link in the show notes. The full transcript of today’s conversation will be posted at ltoshow.com, where you can also find our full episode library and LTO Resource Center.

Thanks for listening to the LTO Show. We’ll see you next time.