Data Protection & AI Summit | Eric Herzog, Infinidat

Clips
News
More from Data Protection & AI Summit

Christophe Bertrand

Principal Analyst

SiliconANGLE & theCUBE

Eric Herzog

CMO

Infinidat

AI and ML drive Infinidat’s approach to enterprise storage performance and protection

Is there a place for artificial intelligence and machine learning in today’s enterprise storage technology? Infinidat Ltd.’s data storage system offerings demonstrate how AI/ML can be leveraged for primary storage at scale.Eric Herzog, CMO at Infinidat, talks with theCUBE about enterprise storage.Infinidat employs Neural Cache, a patented technology driven by AI and machine learning techniques, to intelligently place data in a tiered storage design. It is a process the company will soon enhance, according to Eric Herzog (pictured), chief marketing officer of Infinidat.“When you use a flash array, as you fill it, it slows down,” Herzog explained

play_circle_outline Enhancing Operational Efficiency: InfiniBox AI-Powered Storage Solutions for Smarter Data Management and Optimization

play_circle_outline Collaboration with Index Engines enhances cyber detection capabilities using AI technology.

play_circle_outline Strengthening AI Compliance: The Shift to Private Clouds and Data Repatriation for Enhanced Security and Governance

play_circle_outline Empowering IT Professionals: Essential Education on AI's Impact and Cybersecurity Strategies for Data Protection and Storage Solutions

play_circle_outline Data protection and AI integration leads to next-generation data security solutions.

play_circle_outline Businesses must consider data integrity to avoid operational disruptions and losses.

Info
Transcript

Eric Herzog, Infinidat

Christophe Bertrand

Principal Analyst SiliconANGLE & theCUBE

HOST

Eric Herzog

CMO Infinidat

Eric Herzog, chief marketing officer of Infinidat Ltd., joins theCUBE’s Christophe Bertrand at the Data Protection & AI Summit to explore how artificial intelligence is reshaping modern data protection strategies. With deep expertise in enterprise storage and security, Herzog offers a behind-the-scenes look at how Infinidat is integrating AI and machine learning into its flagship InfiniBox platform.

The conversation highlights the power of real-time protection, especially for primary storage environments where AI workloads introduce new vulnerabilities... Read more

explore Keep Exploring

What are the benefits and functionalities of the InfiniBox technology in managing data placement? add

What new technology is being introduced on September 9th, and how does it relate to AI and ML in the context of storage and cybersecurity? add

What are the reasons behind the trend of data repatriation in organizations, particularly related to security and the use of proprietary data in AI applications? add

What do surveys reveal about the preparedness of CIOs and CISOs for cyber attacks and their considerations regarding AI in data storage and protection? add

What is the concept of next generation data protection in relation to AI and primary storage? add

What are the key reasons for the importance of on-premises data storage in the context of AI processes and data governance? add

bolt Powered by CUBE AI

Eric Herzog, Infinidat

search

Christophe Bertrand

>> Hello and welcome back to the Data Protection and AI Summit. My name is Christophe Bertrand, Principal Analyst at theCUBE Research. I am joined today by Eric Herzog, is the CMO at Infinidat. Eric, great to have you back.

Eric Herzog

>> Thank you. We love being with theCUBE. You guys do great stuff, not only for vendors such as us, but really what you do to help the end users and the channel partners that also support those end users. theCUBE is very valuable from an end user perspective and what the enterprise is like to hear about from you guys.

Christophe Bertrand

>> Well, thank you. And we have quite a bit to cover a pretty vast topic that we're trying to cover 360 here. So let's talk about data protection and AI. When I say data protection and AI, what does this evoke for you? What does this mean?

Eric Herzog

>> So it means a couple things. A is leveraging AI technology into what I'll call next generation data protection. It's not just about backups anymore, which is what data protection historically has been. It's about primary storage as well, and protecting the data there on primary in real time, not waiting till the backup is there, and also that you need to have a cyber mentality built in. Second thing is AI workloads. The more AI workloads you use, not only are they hard to get going and how do you do it, and there's compliance issues and regulatory and legal issues independent of that, but then the more you're using it, you can't have that compromised just the way you can't have your Oracle database compromised, your SAP, your large file systems. You can't have that AI workload because essentially it may be AI, but it's just as if you have someone running your supply chain and it's a human today and you start switching AI, you're still running supply chain information. So if the AI gets hacked, guess what? It's still wrong and the screws don't go to the right factory at the right time. So that's bad too. So AI workloads shouldn't be, "Oh, well it's AI, we don't need cyber or any... " It's crazy. It's just another workload, just happens to use AI.

Christophe Bertrand

>> Right, absolutely. So that's one of the interesting points that you're making here, and we'll cover both dimensions, I think, in detail. Let's start with the feature set that you can augment with AI, especially in the context of storage. You've done a lot of work in what is ML and AI. Can you walk us through what you've done and maybe be specific about it? How does it help me as an IT professional or storage professional? What do you do that is really different from maybe a few years ago? I come from a background of storage just like you and well, many years ago this would've been a pipe dream and here we are. So tell us more about that.

>> So we use it in two ways.

Eric Herzog

>> On the primary storage, our InfiniBox technology, what we do is we leverage AI and ML to manage data placement. So think of it as the way to tier data and what happens when you do that, it allows you to more effectively use things like a DRAM cache, our neural cache technology is over 20 patents. So what that allows us to do is most time, particularly when you use a flash array, as you fill it slows down. Okay, so what do you do? You have to buy another flash array to run more workloads. With us, that doesn't happen. So that's all done with our capability of using AonML to manage the data placement. This goes on DRAM, this goes on flash in the case of our hybrid, because we still do have a hybrid as well as an all flash. It goes out to that to there. And we have some other technology we'll be introducing on September 9th that will expand that portfolio, again using our neural cache AI and ML technology. Second thing we do is we partner with a company called Index Engines. Our product is called InfiniSafe Cyber Detection. We've done some joint development with them. And what that does is allow you to use AI and ML technology to scan elements of storage. It could be a snapshot, it could be a file system, could be a volume, could be a VMware datastore. So that way you can scan it and it allows you to do two things, one, it could be an early warning system. You're scanning certain snapshots every other day, and when it sees something in InfiniSafe Cyber Detection, you can automatically send messages over to your SIEM or source software or if you have a security operation center. So storage actually can be proactive in helping identify threats. Okay. The second is you've had a threat, so you now need to get recovery and you need to recover quickly. So we cover a quick recovery we have, immutable snapshots, we can recover primary storage in a minute or less, no matter how big it could be 4, 5, 8, 10 petabytes in a minute or less. But you still have to make sure it's a known good copy where InfiniSafe Cyber Detection with our partner Index Engines allows you to take candidates, put it in a fence forensic environment, and this is after an attack now and scan. So in this case, the AI would be used as an early warning system, or if you don't do that, and that's okay, you pick what you want to do with the software. You could use Cyber Detection and use it as the capability of scanning in a fence forensic environment. Otherwise, you have to do it the old way, which is you call up the Oracle guys, the SAP guys, "Could you take a look at this? Let's see if there's malware around somewhere. " And that's much more time- consuming than using the AonML technology in InfiniSafe Cyber Detection with our partner Index Engines. Way more time-consuming if you do it the old-fashioned way.

Christophe Bertrand

>> And actually we have a great conversation lined up with Index Engines, so we'll talk about that in this context. So let's double click on a couple of things here. Definitely two sides of the same coin. Look, everybody's saying AI is going to be creating a lot of data. There's no question about that. It's going to generate its own. So I'm sure you're seeing already, in some of the use cases you are dealing with, a lot of organizations deploying a lot more data for AI. Do you think today that that data is actually safe and usable for AI? Are they doing the right thing? And what do you do to really help? Because we're talking about something here that's at scale and will grow overall in this market in the next five years to hundreds of zettabytes potentially of new data. Where's that all going to go? How are you going to manage that? How do you make it safe?

Eric Herzog

>> So I would view AI as just another type of workload. It's another type of application software. It's finance. Oh, you're not using Oracle or maybe using AI to manage the Oracle or SAP or your SQL databases and then run the right financial analysis on accounts payable, accounts receivable, all the stuff you have to do. Your supply chain, same thing, it's a database. And by the way, that database often is also cross-referenced with physical files to show, the screw, the bolt, the nut, the this, the that, so that you order the right things, right? So AI is just another workload. So it is subject to attack the same way the Oracle database is, the same way any of your snapshots are, the same way your volumes are, your file systems. There's no difference. It's still data. So while AI is running it versus a human running that data in Oracle environment, you have AI doing that work instead, you're still running the data in that Oracle environment to get the right screws, the nuts, and bolts to all your factories all over the world. So the only difference is using AI to automate, think of AI as true automation and maybe making some decisions that an automation software would present to an end user and say, "Here's the three choices, I think the screws need to go to Paducah, Kentucky, these ones need to go to Tokyo, and these ones need to go to Johannesburg. " So how you do that, the human recommends it, or AI recommends it for you. But that means if the data is compromised, it's the same thing. It's still that Oracle database that AI is running instead of the human. So that's what matters is making sure that AI is just another workload and the data that goes into that workload is subject to attack, just like the data going into your SAP environment or into any other environment you're managing as an enterprise IT person, or of course the enterprise security team needs to be concerned about. It's the same thing, just AI is the incredible amount of automation and some intelligence around that automation. That's all it is. It is not immune to the attack. Data is data.

Christophe Bertrand

>> Absolutely. That's really the main point here. That's why we have this summit to talk about how you protect data in this age of AI, how you leverage AI as a friend and defend it instead when it's a foe, when it's actually AI powered cyber attacks. Couple of things, we haven't talked about compliance and governance yet. I think governance is a bigger topic, but just when you think about compliance, earlier, we were talking about the number of regulations that exist today, just in general terms. There are lots of them, and now there are some specific also AI regulations. Now in reality, I expect that a lot of the AI workloads, to use your term, applications workloads, will not be in the cloud that for some maybe legal, maybe compliance, maybe governance, some organizations and many enterprises will actually run on-prem, which is really where you come in. Now, all they could run in a private cloud type of scenario, but storage is going to become more and more critical, especially on-prem in a very weird twist of history here. What's your take on that? Do you see that already happening? What do you think?

Eric Herzog

>> We see what's often referred to in the IT press as well as the storage press, as data repatriation. So first of all, from an AI perspective, there's lots of concerns. If you're using chatbots that are open source, if you use that and you're publicly traded, you just exposed your IP. Let's say I'm not, I'm a little tiny startup, I got these wiz kids developing the newest Hinkelmeyer and we're going to rule the world, and we're going to make billions and our investors are going to make billions. Well, if they use an open source chatbot, guess what? All the big competitors just figure out how did you Hinkelmeyer. Because we told them because legally they can access that data. So the reason for data repatriation are things built around security, in the AI world, being able to use your proprietary data confidentially. And by the way, private cloud is basically a private cloud. It's the version of Oracle Cloud or Amazon Cloud or Google Cloud. That's your own, right? And you don't expose it publicly. So if you're using that way, you want to use your own proprietary data for AI workloads. So that is a huge, huge issue. So more and more people will start doing AI as all this legal and regulatory and compliance issues get resolved. But we see customers, our customers, because we mostly sell the global Fortune 1000, we see a lot of our customers already playing with it, but they're not using an open market chatbot, right? They are now LLMs and SLMs, vector databases that you can get from software companies that are just for you, your company, not open source like the chat GPT, which is open source, or you do it yourself, right? Some of the larger, the Fortune 200, they got a lot of software guys. So they may be developing their own and then they use their own proprietary data to get sure the supply chain is right or the finance is right, whatever application they're using, that it is the right data and it's proprietary to them. They don't have to worry as much about hallucinations and they clearly don't have to worry about the legal issues of using open source AI technology, which means you just could steal stuff without really realizing you're stealing it. Also, someone could steal your stuff without you knowing that you really just exposed it. So AI is really, the takeoff of AI hasn't hit yet, as the compliance regulatory and legal issues get done, data repatriation or using it in the private cloud will get even more important because you don't want this data exposed. And so from a storage guy perspective, it's great and obviously have our own AI RAG technology, a retrieval augmented generation, which works not only with our own storage, but actually works with our competitors as long as it exposes an NFS protocol, which all the databases do, and obviously all the files do. So you could use our storage in a competitor's storage with our AI RAG to get that right proprietary data. But you have to make sure you understand what you're doing with the data, why you're doing it that way, and again, several times I've been to some events where literally there's all day session on AI and over half of it is a bunch of attorneys and regulators talking to a bunch of IT guys about, "Okay, blah, blah, blah. If you do this, someone will sue you. If you do that, you're giving your IP away for free. " And IP isn't just in high-tech. If I develop the best running shoe in history of the world, it may be only a running shoe, but if it's the best in the world, I'm going to make billions, right? I mean, that's what Nike did, right? So I don't want people to steal that, and I have designers that use computers to design running shoes or tennis shoes or tennis rack, anything, cars, whatever. So it is a real issue of protecting your IP.

Christophe Bertrand

>> Right. And this is really where we talk in a broader term of data protection, not just protecting from destruction, but actually there is that risk too, because what you've described makes perfect sense. I think we'll talk more about this in our governance summit in the edge of AI in the September timeframe with my colleague, Scott Hebner. We will cover all of these topics and regulations, which will make for an interesting time. We'll make it interesting, not boring it can be. But the point is this, the thing you've just touched upon very important topics here. The one thing that strikes me is that it doesn't appear to me that people are thinking as they are building this data infrastructure about protecting the data the way they should, not just from leaking or being poorly used, that's a big issue, not arguing that point, it feels to me like the fundamental work, just like any workload, has to be the protection, the protection in the context of recoverability, the protection in the context of immutability. We're hearing every day now of prompt injections where people go essentially attack the AI software itself, potentially causing a risk of data destruction or all sorts of other issues which are so far-reaching. We also live in a world where geopolitically, things are not that great. There are lots of attackers from areas that are also being physically attacked and physically attacking others. So all of this is starting to combine into placing the storage and the data in first line of attack. So with that in mind, do you think that today your clients understand that AI by design is actually data that's protected by design? Are we there yet? Is there a lot of education to do?

Eric Herzog

>> So there's still a lot of education to do. If you look at even regular storage, regular data, and you look at surveys of CIOs and CISOs, they freely admit in many of the surveys that are out there from industry analyst firms or sometimes from the financial press, they're not really prepared for a cyber attack and they haven't even thought about AI because it's new. So if you think of AI, like it's Oracle, it's SAP, it's going to be on your primary storage. The next generation of data protection is where the difference comes in. It's about protecting data on primary, AI's on primary. By the way, you still should be backing up your AI workload when it's done, right, you're running it constantly. So great, you have to back up the fact that you're supposed to ship the screws because you just did a run of your supply chain and you do runs of supply chain big companies every single day. Well, guess what? If it was done with Oracle, they back up the Oracle database. So now you back up the AI. So that's okay, that's the data protection in the traditional way, but you want to be protecting it in real time while it's happening on the primary side. So the next generation data protection is about protecting data primarily from cyber with a cyber recovery angle to it for a traditional backup workloads, but also for primary storage, making it immutable, scanning it on a regular basis, which is what our InfiniSafe Cyber Detection does, is scan on a regular basis, making sure you separate the management plane and the data plane, making sure that if you do have an attack, you can create a fenced environment because you need to find a known good copy before you do that recovery, you want to make sure that you have incredible recovery times. We guarantee SLAs on primary storage and on secondary backup. On primary storage, we will recover no matter how big the dataset size is. It could be four petabytes, 10 petabytes, 20 petabytes, one minute or less. And in fact, because I'm old school, I believe that if you can't demo your software live, your software is not very good. And when I started in this business, you demoed it live at a trade show or at a lunch and learn or at a seminar. No one was videoing it and then showing it. Don't do that. It was bad. So guess what? I still do it that way. I make my teams do it live. We recovered four petabytes in four seconds live, four petabytes in four seconds on primary. You also want to be able to have rapid recovery on secondary because what if there's no good copies on the primary side, you're no good copy. You do find in a backup repository. So you've got to bring that repository available to click the button so that Commvault, Veeam, Veritas, IBM Protect, Zerto, whatever is your backup software choice can actually do the recovery, right? So you've got to do it both ways. In that case, we guarantee the recovery in 20 minutes or less. And when we did the live webinar, because I'm a crazy old geese bag, guess what we did? 25 and a half petabytes in 12 minutes. Now that is rapid. So you've got to make sure that that is built in. And as I mentioned earlier, a lot of the CISOs or the CIOs don't think of storage as part of a comprehensive cybersecurity strategy. So you want to make sure that you do that. You're protecting the edge, you're protecting the servers, you're protecting the apps, you're using your SIEM or source software, you've got a SOC. Remember the days whenever we used to have a knock when we were younger, when I had brown hair? Now they have SOCs, Security Operations Center. That's what they do these days. So you want to make sure that that is all monitoring and you want to make sure that the storage is part of that component. Whether the SIEM or source software tells the storage, "Hey, I sense a threat. " And the storage starts doing things immediately to shrink down that threat window, or the storage is scanning, let's say, once a week, every twice a week, and senses something and then sends a note over to the SOC and says, "Hey, these two storage arrays on these snapshots, something is flaky. " And with our cyber detection, which we do with our partner, Index Engine, it's 99. 99% validated by analysts as accurate. So it could be a false positive, but it's pretty low. So instead of being reactive, which by the way, InfiniSafe Cyber Detection do, put in a fence forensic environment and start a scan after you've been attacked. But the other way is as an early warning system. So storage can be part of a very comprehensive cybersecurity strategy, clearly for recovery, both the speed and scanning, but you can also be proactive and communicate with SIEMs or SOC if you scan on a regular basis.

Christophe Bertrand

>> And by the way, what you're describing, and we'll cover that more with Index Engines, is actually AI powered.

Eric Herzog

>> Yes. - So this is AI coming to the rescue

Christophe Bertrand

>> of AI infrastructure attacked by AI powered cyber attackers. So that's why it felt so important to cover this topic of AI and data protection to me, because there is so much going on at the same time. It is very confusing. As I said a couple of times, I think we're seeing the convergence of multiple markets and technologies that are really changing the way organizations organize from an IT and security standpoint, but also a data management standpoint. We're seeing this new workload called AI, which is massive, potentially, will be life changing, will certainly augment a lot of processes, will also make decision or decisions on its own, I should say, through agents. Agentic AI is a thing. It will happen. It's happening. Now, is it happening enough? Probably not to show true good ROI, but it will soon. It's only a matter of time. What I like about our discussion is I think you've really positioned the fact that data and data that is on storage on-prem is actually fundamental. It's fundamental to AI processes. It's fundamental to the protection of the data that allows you to run AI processes. And I think people are going to realize that it is fundamental to their governance and to limit their business risk. Now, I'm not saying there won't be any cloud- based solutions as well. Absolutely, it's going to be hybrid. We know that's the way of the world. But I do believe for all the reasons you've mentioned, there is going to be a resurgence of on-prem, at- scale storage. If I were to summarize, what I heard is, and it's very important for the protection and the recovery piece of data, the fact that you can do things in seconds with very smart snapshots and great use of placement of data and what you described earlier, some AI in there as well, changes the game. The fact that you can do this constant testing with the help of partners like Index Engines is also very key change the game. It makes AI possible. And it sounds like it should be obvious because it's about the data, but I think a lot of people are looking at the shiny objects, but they're not looking at the fundamental infrastructure, which is what I think you've described really well.

Eric Herzog

>> Again, garbage in, garbage out. So a human running Oracle databases, if the data coming in is garbage, it's data coming out with a human being standing there. So if you have an AI agent doing the work and it's garbage in the data, garbage out, it's still garbage. So an AI hallucination, quite honestly, is like a human making a mistake. In my mind, it's just the same. So you have to realize that AI is, if you will, automation on steroids. Really automating it. Incredibly automating and allowing that automation to make some decisions on its own. Okay, great. But again, if the data underneath that automation is lousy, then the result is lousy.

Christophe Bertrand

>> And what if there's no data? What if somebody, again-

Eric Herzog

>> Ransomware or malware? Then you just totally ruined it. And if you're using it, for example, for supply chain, you're not shipping and the factory goes down. So you're thinking, "Oh, my data center's down. I don't like it. But yeah, I get it that up in a day or two. " But all the factories are down. They don't have screws, they don't have tires, they don't have hubcaps, all the stuff you need if you're a car manufacturer or if you're making drugs or whatever you're making, you could be making, again, paper products. You have to get the paper, right, you got to put it in the roll. Then there's always the thing in the middle, right? For your paper towels. Then they have to put it into packaging. Then if you get the big box that's packaged up when you go to the grocery store, that all is done robotically and all done by computer. So if you don't have the right wrap there, you've got all of those paper towels just sitting there because you can't ship them as just paper towels. You have to put them in the plastic wrap and then put them in packaging to get to the grocery store or to Costco or wherever you're trying to get it. And so AI just is automating that process. If the AI data is attacked, the only thing it's done is you get to say it's AI. It's the same problems if it was attacked with a human deciding when to get the plastic over to the paper factory, it's the same thing. Just AI now is messed up too, just as if a human was. So it's all about the data. The core is the data, which is why you need next generation data protection on your primary storage, which is what AI would use. And you need next generation data protection on your backup as well, just in case you can't recover from primary, you got to recover from backup. I need to make sure that they're clean using AI and ML technology with our cyber protection technology that we've done jointly within Index Engines.

Christophe Bertrand

>> Eric, what a great summary. I could not have done it better. So I'd like to thank you very much for your time today. This was great. I think we covered a lot of ground, and obviously, I want our viewers to also think about taking a look at this great session we have with the Index Engines as well.

Eric Herzog

>> Yeah, that is going to be very good because it delves deeper into using AI for the good guys versus AI for the bad guys.

Christophe Bertrand

>> Exactly. - And the bad guys use AI to, trust me,

Eric Herzog

>> everyone's using AI, right? Everyone uses a pen. Everyone uses computers, including bad guys. So the reality is that session is going to talk about using AI for good, to defend your data, defend your enterprise, versus what the bad guys are trying to do with the exact same, by the way, technology.

Christophe Bertrand

>> Perfect. And I can assure our viewers, we are real. The shirt is real.

Eric Herzog

>> Yes. - It is by design.

Eric Herzog

>> We are not AI generated.

Christophe Bertrand

>> We are not AI generated. They couldn't do that.

Christophe Bertrand

>> Well, Eric, thank you so much.

Eric Herzog

>> Great. Thank you. Love being with theCUBE and appreciate all the stuff you do, not only for us, but really our end users have made some really nice comments to us about, we saw this on theCUBE. They gave us real information, particularly due the summits where you have a whole bunch of stuff, when you do just the one-offs, like sometimes you guys do at the trade shows, those are nice, but you have all the, in this case, you have a focused summit topic, and customers, they see one thing, then another, and they're all really that key topic. Customers love that. That's what they've told us.

Christophe Bertrand

>> Thank you so much. I'm going to hire you as my agent for promotions. Thank you, Eric.

Eric Herzog

>> Thank you. - And to our viewers, stay tuned.

Christophe Bertrand

>> There's more coming up.