In this episode of Infrastructure Matters – Insider Edition, Randy Kerns is joined by Brian Pawlowski, Chief Development Officer at Quantum, for a conversation on Quantum’s Myriad solution and how it is uniquely offering solutions to customer data challenges.
Their discussion covers:
- Quantum’s view on the current market and how storage and application have evolved, including customer use cases
- The challenges facing customers within storage and data management systems, today
- Quantum’s current product portfolio, including how their latest solution, Myriad, fits into their portfolio
- How Myriad offers unique solutions to help solve data challenges, and where Quantum plans to take Myriad next
Learn more about Myriad Software-Defined All-Flash Storage Platform for the Enterprise on the Quantum website.
You can watch the video of our conversation below, and be sure to visit our YouTube Channel and subscribe so you don’t miss an episode.
Or listen to the audio here:
Or grab the audio on your favorite audio platform below:
Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this webcast. The author does not hold any equity positions with any company mentioned in this webcast.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.
Transcript:
Randy Kerns: Hi, this is Randy Kerns with The Futurum Group, and we’re going to do an Infrastructure Matters session with Brian Pawlowski of Quantum, or in this case we call him BP. BP, it’s great to see you again. We were just talking about this. We saw each other at the Flash Memory Summit. I thought this would be a great chance to catch up with you and get your perspective. You and I have both been in the industry for a very long time, so it’s great to kind of touch base and figure out what’s going on. We’ve got a few questions here that I think will be good prompters here. What do you think has been changing in the storage and applications market compared to, say, 10 years ago?
Brian Pawlowski: Flash one, nothing is replaced tape, and disc is squeezed in the middle and there’s no pleasant outcome for disc. I think what really has changed is, and it’s obvious, is just the emergence of flash and the growth that few people predicted back in 2010, 2012 predicted that we were going to be this far this fast with the deployment of flash. I think that’s sort of, to me, the most interesting part, right?
Randy Kerns: Mm-hmm.
Brian Pawlowski: But coupled to that, regardless of technology, is data explosion continues unabated. This is the thing, when I started at NetApp, we were really excited. We were going from a one-gigabyte drive to a two-gigabyte drive and seven drives in the rate group to 14. We were going for the big time for storage deployments, and I was back in ’94. Nobody thinks in terms of those sizes anymore for anything, not even for flash, right? I think you’d struggle to buy a USB stick that size, right? I think that the size of systems and the capacity of the systems and the explosion of data growth has just has gone in unanticipated ways. I just don’t think anybody would predict where we are right now. They would be low balling it far below where we are.
Randy Kerns: I absolutely agree, and it’s stunning to me when I reflect back on over my career, how things have changed. There are people that say storage is boring, but they’re not paying attention because it dramatically changes all the time and there’s new developments. It’s pretty exciting to me. But the big issues I see are in managing all this information, so what do you see as the limitations today for that?
Brian Pawlowski: The limitations for the management of data is, it is the needle in a haystack problem. You know you have it somewhere. You know the information is there because you basically decided that your move-forward strategy is going to be save everything, because God knows I might need it for my next AI application. I was just at Carnegie Mellon University at their Parallel Data Lab retreat, and several of the researchers there, the industry researchers, were arguing about, not a really smart approach to save everything. You kind of have to curate it. We were arguing back and forth about it. But the thing is basically finding data, finding the data you want quickly and not just by file name. You start looking at object, you start looking at video. t my current employer, we have a media asset manager product and that allows you to basically, I call it deep packet inspection of data, that allows you to find bits of data by facial recognition, by an object in the frame, in an hour long video and to tag it and go back there anytime you want. So it’s no longer find a file. It’s find the interesting bit of the content you need at that moment in time as fast as possible to make use of it for a lot of industries. So anyway, so it’s gotten a lot more complicated and the pile you have to grope through typically is much, much larger now.
Randy Kerns: Well, and it is certainly okay to talk about your current employer and the product here and name that because that’s actually an outstanding solution I’ve worked with for a number of years. Let’s talk about the new product Myriad and what it brings to the market and what’s different about it. I’ve been pretty excited about when it was initially shown to us. Tell me a little bit more about it.
Brian Pawlowski: So in the simplest sense, and that’s not going to get anybody excited, Myriad is an all-flash array. It’s basically designed to work fast. Second level is, Myriad is a scale-out, all-flash array that can grow to any size you want for as much performance and data scaling as you need. I think the place where it gets very interesting is, Myriad is primarily a storage software solution built in terms of containerized technologies to support large systems scale out in a very uniform way. I really do think it’s a cookie cutter approach. It basically makes no assumptions about the hardware except basically three things. One, you have NVMe flash; two, you have an NVMe fabric; and three, you just have a CPU. Now what didn’t I mention in there? There’s no requirement for HA. You don’t have to have dual ported SSDs that lowers the price.
You don’t have to have specialized server hardware, use compute nodes. So, truly, truly commodity hardware. Honest to God, I’ve said so many times in every company I’ve been in that we use commodity hardware, but this is really off-the-shelf hardware that we couldn’t have made a product of except everything basically evolved to this point. NVMe became a standard.
And NVMe is a great enabler for flash performance. NVMe fabric basically allowed compute nodes, CPU complexes to talk directly to the storage in a crossbar situation, and we have an extremely uniform cookie cutter scale-out solution that you said before what happened 10 years ago. 10 years ago, NVMe fabric didn’t exist. 10 years ago, Kubernetes didn’t exist. And now we’re using basically a mini cube kind of cluster based on Kubernetes technology and a container architecture all the way through that allows us to sort of build it out of logical building blocks very quickly. Anyway, so we’re taking advantage.
Randy Kerns: Well, one of the things, and you brought it up, is this east-west traffic thing has always been a limiting factor. Now you’ve seen the way to overcome this or compensate for a great deal of it, so that’s very important. I agree, absolutely. So tell me, how’s this fit into your bigger picture portfolio that Quantum has?
Brian Pawlowski: When I got to Quantum, we basically cover a lot of, we have seven products. We span the spectrum from capture with video surveillance, and as Azure aside, I keep hearing that … First of all, when we talk about the data explosion, it’s unstructured data. We’re not talking about database kind of stuff, but 60% of the unstructured data comes from video surveillance. I mean, we’re on cameras now, but I mean video surveillance includes body cams, right? Video surveillance is in every 7-Eleven you go to. Banks, ATM machines, every street in London probably, just nonstop. So that’s a gigantic part of the unstructured data explosion. So we have that. We go to our StorNext File System, which is a high-speed video post-production file system that’s been used by most of the major studios worldwide and sports teams.
By the way, when we say video, you think studios, but now it’s sports teams, franchises, all the networks that cover them, and corporate video. Everything’s corporate video these days. In fact, I was trying to figure out how to fix my camera this morning before we got on, and I was getting YouTube video suggestions that I’m like, “I don’t have time to watch a YouTube video. I just want to see the next few steps I have to do.” We also have ActiveScale object storage, and this is another thing that’s really exploded over the past 10 years, is basically S3 as a standard, pushed by Amazon, and it’s a standard for data access in the cloud and on-prem. And then we go into our tape robots and our automation, we call them automation. These are the gigantic robotic tape libraries that basically store lots of data cheaply. What we were missing was that the thing in the middle for primary storage, for general purpose, high-performance applications, we had the ingest part. We had a specialized video post-production file system object storage, which is sort of like the next middle tier over and then tape, and Myriad basically plugs the hole in the center and provides a really high performance solution for what I would call modern data applications.
Randy Kerns: And it can scale very large, which is kind of where we started this discussion. One of the things I do want to maybe reflect on is, I spend a lot of time with our large client, IT clients, and it’s somewhat of a defense mechanism in that the IT people keep all the data because they don’t want to fight with legal or anybody else about what needs to be deleted. So to keep everything is the lowest common denominator of not having to battle for saying you don’t need that anymore. And that has changed everything from the standpoint of lots of data but being kept forever. So now you end up with these data management and migration or data movement capabilities, like you have a StorNext, becoming so valuable and now you’ve filled this giant hole in the middle. This is great.
Brian Pawlowski: Yeah. You saw my data factory picture.
Randy Kerns: Mm-hmm.
Brian Pawlowski: I think that’s the thing. My colleague Eric came up with this model of data as something that’s transformed and manufactured into a product just like a car. I like the blue collar kind of analogy for this stuff, because we always think airy thoughts, right? You have raw materials, which are the capture points like video surveillance or medical imagery, and then you put it into the processing, which is where we intersect with AI today. That’s becoming so much a part of the transformation of the data to something usable, which is where I was talking about facial recognition and things and the AI part. And then you have the finished goods, and then you go store the finished goods. Where do you store the finished goods? And you’re going to store it on lower cost storage. You put it on the retail shelf for sale. Right? The thing about data where that factory model breaks is that data’s reusable. You pull it off the shelf, you put it back into the process, you redo it and make it better and stuff. It’s fascinating that the data moves all through the … I don’t want to use the word tiers because that kind of bothers me a little bit, because it’s so fluid between cost points, where it’s being stored based on the performance required at a given time and how much you want to pay for it because of how it’s being used at that moment. But it moves back and forth. So go back to your finding the data question: where?
In all of this kind of circular pipeline, that also includes the cloud. All of our products basically interface with the cloud and can use the cloud as a final destination and then pull it back from that, maintaining the catalog and location of that data and then go back to … I want to find, I don’t know, I want to find all of the video clips of Billie Eilish in the LA concert in December. I went there, by the way, in December, Billie Eilish conference when she was whatever, the night she played with David Grohl. Okay. So I need those things because I’m doing a news thing later on. Anyway, finding that data across all of these storage solutions, that’s the complexity of what we’re dealing with today. It’s not just one storage system anymore, right?
Randy Kerns: Yeah. There’s one more thing observational I’ll make, and I’m sure you didn’t plan this but your timing is great from the standpoint that you’ve got a high performance NAS system effectively, and now we’re at this point with AI and needing high performance access to lots of file data as more and more people want to use private data to train or tune these foundational models. This is going to be a solution that you’ll push into that market and be very useful there.
Brian Pawlowski: Yeah, no, I agree. The repatriation from the cloud is real. I think we’re reaching a point of pragmatic cloud usage. I was on some call with some analyst, I don’t know, I’m not sure when, in 2009, 2010. They were like, “Everything’s moving to the cloud.” And the cloud at that point was, the only cloud was Amazon. Everything’s moving to Amazon. Then you got, fast-forward six, seven years, and it’s like, which cloud are you moving to? How many clouds are there? And now people are looking at the problem in a more complex way, and maybe hybrid cloud is the word I want to use here, but they’re going to use on-prem, they’re going to use cloud. Data’s going to move back and forth. Cloud is not the final destination anymore. Going back to the data movement, you bring in the data, having the data where it needs to be, when it needs to be, and at the cost it deserves to be, right? So it moves back and forth. So it’s gotten interesting, I think. Going back to your comment of the data, finding the data using the data observation that you made, that’s the thing to kind of solve. Right?
Randy Kerns: Yeah.
Brian Pawlowski: It makes it all very interesting. It’s not a storage problem per se, it’s a data problem, right? My diagram did show all the storage infrastructure in the basement, like the plumbing.
Randy Kerns: I have these conversations with a lot of clients about data versus information, data, bits of ones and zeros that you organize and everything. It’s information when you know what it is.
Brian Pawlowski: Right, right.
Randy Kerns: And that tends to ring home with them because that’s the real problem in that they’ve got lots of data, but there’s an awful lot. They’re not really sure what it is. And that plays exactly to what you’re saying.
Brian Pawlowski: That’s an interesting … I haven’t thought about distinguishing data from information. I always think of data as the user stuff and storage as the stuff that I work on, but I think information has meaning of … What is it?
Randy Kerns: Context.
Brian Pawlowski: Meaning. Right. And I think there’s a metric for information meaning density or something like that for spewing lots of data and not having any information whatsoever, or a very information dense kind of data. That’s an interesting way to look at it. Yeah.
Randy Kerns: All right, well, this has been great conversation. Of course, you and I have known each other and could go on for hours here. However, we’re somewhat limited, so we probably need to close out. Any last thoughts about where we’re at and what’s going on and what the future’s going to be, especially around where Myriad’s going to take you?
Brian Pawlowski: Yeah, so we did Myriad. One of the goals was, like I said, to avoid any specific hardware requirement other than what’s available off the shelf. Truly off the shelf. Just like no special requirements whatsoever. So the software is hardly, it just runs on compute nodes. If we go grab an OCP box, we can probably run a Myriad on it without a problem. So we’re looking at the future of Myriad as primarily being just software running on different people’s hardware. And again, you’re going to qualify this stuff, but we’ve already run Myriad in the cloud as a proof of concept because it’s just software. So we’re going to be releasing cloud versions of Myriad to give you kind of a fluid storage solution that you can manage from across the whole spectrum of your deployment.
Randy Kerns: Common experience.
Brian Pawlowski: Yep. Yeah.
Randy Kerns: All right.
Brian Pawlowski: Can I say one more thing?
Randy Kerns: Absolutely.
Brian Pawlowski: So the talk I was asked to give at the Flash Memory Summit was, can you come in and explain to us about tape and ActiveScale Cold Storage, kind of the Glacier approach of object-based tape storage? I mean, you asked what’s changed in 10 years, what’s not changed? It’s funny, I cannot get over it, but tape is still a mainstay for the durable long-term storage of data at the lowest cost possible because you put the data on the tape and you pretty much put it on a shelf. It draws no power, it sits there pleasantly, and you pull it back and you read it back at some point in the future. It’s a well-known technology. I had mentioned at the Flash Memory Summit, we are looking at finding the replacement for tape because everybody asked that question, “Well, what replaces tape?” I think my last line at the Flash Memory Summit was, “Well, nothing yet. So you better love your tape now, hug your tape because it’s going to be once in a while.” But I think that’s going to be kind of a fascinating place, and that’s probably the most boring. No offense to all the tape heads out there. If there’s something, if storage is boring, flash is really boring. Sorry, tape is really boring, and yet I think it’s going to be really important in that space because where’s all that data going to go? You can’t afford to shorten the most expensive parts of the system and you want it for a long time. So you need a durable, long-term storage option for low-cost storage. I oddly expect there’s going to be more innovation there. I witnessed Amazon’s Glacier. Amazon’s Glacier is object front-ended, tape-like stuff up there. They don’t talk much about it. Azure has it too. And those storage solutions are to allow people to afford to keep all this data that they feel they need, right?
Randy Kerns: Okay. So let’s talk for a second about ActiveScale with the ability to tear off the tape. How does that all work with Myriad today? What’s the process the way data would flow here?
Brian Pawlowski: So Myriad is, as you mentioned, basically a file and object system. It’s a high performance solution, and it’s all-flash base so it sits at the primary storage part of the pipeline, right? ActiveScale Cold Storage uses active scale or S3-based object storage system with tape backend, and it essentially provides an object storage interface to tape. It’s that simple. So we’ve been in our product line basically using S3 as our, I keep using the word tiering, I got to use it this time, tiering solution as the standard interface to tier to tape through ActiveScale or to tier to cloud. Our customers have this in production with StorNext. So our file systems, automatically, we can set up policies, but this is not something you do manually. This is the scale of data problem. You set up a policy of when to move data off of your active part of your system, very expensive, which is the flash system to your tape. So when I look at ActiveScale Cold Storage, I look at the S3 object front end of ActiveScale as the translator to tape. No one is using tape. Our customers are increasingly not using tape as tape. They’re not backed up as-
Randy Kerns: Back repository.
Brian Pawlowski: Yeah. This is basically taking files and blobbing them up and moving them to ActiveScale, except it’s basically moving it to tape. We do all the management of that. Or you can move it to cloud. But the important part is that, the two-pole thing of tape and flash, we started with this a bit. Tape and flash, tape and flash, these are the two things that are going to be the important parts of your infrastructure in the future. And then basically moving the data back and forth between them. It’s the policy-based automated movement of that data that’s critical because, as we’ve been talking about, the sheer amount of data we’re talking about in dealing with here, it defies any manual intervention on the part of administrators.
Randy Kerns: So the customers would define what that policy is and set it to auto mode, and you don’t worry about it after that. Right?
Brian Pawlowski: Yeah. We had talked about enhanced deep packet data inspection, so to speak. So this is not just by file name or expiration or modified time or user. There’s other enhanced metadata that we’re tagging the files with in Myriad to allow you to do more fine-grained data motion of when to move data off of myriad onto ActiveScale Object Storage.
Randy Kerns: Yeah. Some of the dynamics about the technology change and forward migrations and all that are considerations that have kind of moved away from the normal IT operations and moved into cloud providers are very, very large organizations. So that fits there, but there’s other dynamics. So it’s not a simple answer anymore as anything. All right, well, this has been wonderful to talk to you again and hopefully it’ll be before the next Flash Memory Summit where we can get together.
Brian Pawlowski: I hope so.
Randy Kerns: Thanks alot for taking the time, BP, thank you very much. I look forward to the next time we get to talk.
Brian Pawlowski: Okay. Thanks so much.
Author Information
Randy draws from over 35 years of experience in helping storage companies design and develop products. As a partner at Evaluator Group and now The Futurum Group, he spends much of his time advising IT end-user clients on architectures and acquisitions.
Previously, Randy was Vice President of Storage and Planning at Sun Microsystems. He also developed disk and tape systems for the mainframe attachment at IBM, StorageTek, and two startup companies. Randy also designed disk systems at Fujitsu and Tandem Computers.
Prior to joining The Futurum Group, Randy served as the CTO for ProStor, where he brought products to market addressing a long-term archive for Information Technology and the Healthcare and Media/Entertainment markets.
He has also written numerous industry articles and papers as an educator and presenter, and he is the author of two books: Planning a Storage Strategy and Information Archiving – Economics and Compliance. The latter is the first book of its kind to explore information archiving in depth. Randy regularly teaches classes on Information Management technologies in the U.S. and Europe.
