This interview examines Oracle data availability and resilience for agentic artificial intelligence, AI, focusing on Oracle AI Database 26AI, platinum and diamond availability tiers, Exadata integration, zero data loss capabilities and Oracle’s multi-cloud strategy for mission-critical workloads. It is recorded at Oracle Data Deep Dive 2026 at the New York Stock Exchange.
Ashish Ray of Oracle is senior vice president, product management. Ray appears with host Dave Vellante of theCUBE Research to unpack Oracle’s approach to mission-critical data infrastructure. They draw on product management experience to explain database-level resilience, built-in security and kernel optimizations that reduce failover times.
Topics include database-level resilience, upgrades to Oracle AI Database 26AI without application changes, Exadata’s role in maximum availability architecture, MAA, and support for agentic AI in enterprise environments. Ray emphasizes resilience and security baked into the database, seamless upgrades without application changes and dramatically reduced failover times—platinum under 20 seconds and diamond near-instant under three seconds. They highlight Exadata as essential to MAA and recommend that C-level leaders prioritize scalable high-performance data infrastructure to reliably support agentic AI and distributed resilience.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Oracle Data Deep Dive NYC. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for Oracle Data Deep Dive NYC
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for Oracle Data Deep Dive NYC.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Oracle Data Deep Dive NYC. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to Oracle Data Deep Dive NYC
Please sign in with LinkedIn to continue to Oracle Data Deep Dive NYC. Signing in with LinkedIn ensures a professional environment.
In this interview from Oracle Data Deep Dive NYC 2026, Ashish Ray, senior vice president of product management at Oracle, joins theCUBE's Dave Vellante to discuss how the transition to agentic AI is raising the bar on mission-critical database availability, resilience and security. Ray explains how Oracle AI Database 26AI delivers high availability and near-zero failover times by default — requiring no application changes from customers. He breaks down Oracle's tiered resilience framework, where the platinum tier reduces failover to roughly 20 seconds and the...Read more
exploreKeep Exploring
How is Oracle addressing availability, security, and support for mission-critical AI deployments — including enhancements to its AI database and new service tiers?add
When migrating from Oracle Database 19c to Oracle AI Database 26AI, will customers get built-in resilience, high availability, performance, and security by default without needing application changes or additional technologies?add
What architectural changes were implemented to achieve the improved failover times (e.g., platinum <20 seconds, diamond ~0–3 seconds), and how are those improvements delivered "without application changes" — do they apply to any applications supported by Oracle Database or only to Oracle Fusion applications?add
Why is Exadata essential for the Platinum (and Diamond) tier—what role does it play in delivering the performance, high availability, and fault tolerance required for those service tiers?add
How is Oracle making "zero data loss" achievable beyond the Zero Data Loss Recovery Appliance—extending that capability across distributed databases, multi‑cloud environments, and customers' entire estates?add
>> Welcome back to Oracle Data Deep Dive NYC. I'm Dave Vellante, and this is theCUBE and NYSC Wired's coverage. We're here at the New York Stock Exchange. The options exchange has closed down, the bell has rung, and we're wrapping up the day. Oracle's been here all week doing deep dives. They just came back from Oracle AI World in London doing these tours around Chicago. And so what we're seeing is that as organizations are deploying AI, going from experimentation into production, mission-critical is becoming much more important for applications. And what Oracle's been doing and talking about all week is raising the bar on availability and security because you got to trust these systems. And we're talking about systems that absolutely can't go down, they can't lose data. And now we have to defend against an entirely new class of adversaries. We heard about this all at RSAC a few weeks ago, whether it's prompt injection or securing AI or even quantum computing. And so Oracle just announced a number of enhancements to its AI database. They've upped the game on the platinum service and even there's a diamond tier availability service, all free if you've got the right infrastructure installed. And designing security in, we've talked about that a lot. And so these are key innovations or key factors to enable innovation in the AI world. Joining me is Ashish Ray, who's the senior vice president of product management at Oracle. He's going to break down what this all means for enterprises. Running mission-critical workloads and systems, and we're going to bring it home. Ashish, good to see you. Thanks for coming in.
Ashish Ray
>> Great to be here, Dave.
Dave Vellante
>> Okay. So we talked earlier to Juan about these different tiers. The platinum tier, you got the diamond tier for even greater visibility or availability, which folks like the New York Stock Exchange are very interested in.
Ashish Ray
>> Right.
Dave Vellante
>> How is this change... I mean, I remember the days of just where three site was the sort of gold standard. We've gone so far beyond that. How should senior IT leaders think about mapping these different tiers to the different classes of applications and workloads?
Ashish Ray
>> Yeah, absolutely. That's a great question, Dave. So what is happening is so far, or if you roll the clock back, say, just say 10 years or seven, eight years, HA used to be like, "Okay, let's make sure that server does not go down. And please, please, please, let's not lose any data." Okay? So that has been the traditional paradigm so far. What has happened very rapidly over the last couple of years, if you look at the way agentic AI is coming in the enterprise, and these are autonomous tasks which absolutely cannot deal with any bottlenecks, otherwise latencies queue up, transactions queue up. There's also the issue of cybersecurity threats, which means immunity has to be baked inside the database. It cannot be built in the perimeter. And then also we are looking at distributed resilience. So when IT leaders, IT directors, senior technology leaders, when they have to look at resilience for the enterprise, they have to consider all of this and they have to see how much, to what extent resilience is built inside the data. And that's where a big differentiation comes in when it's Oracle.
Dave Vellante
>> I'm glad you're here because one of the things we haven't talked today, and I remember reading the press release about this. So you get the platinum tier, and you can be used with any workload. Again, you got to have the infrastructure in place, most customers do. But something we haven't talked about that I want you to address is, this could be done without any application changes. I read that in the press releases. Because in my mind I was like, "Oh, what are the prerequisites, what do I have to change, and what does it cost?" It doesn't require application changes and there's no charge. Just to upgrade the system, so they got to do that. So what does that say? I mean, we were talking about, earlier I was mentioning three site data centers, extremely expensive, a major architectural change to an organization, but now you're basically saying it just comes with the service. What does this say about your sort of long-term strategy for making resilience a fundamental capability?
Ashish Ray
>> Yes. Yes, absolutely. So that's what I mentioned earlier, that this resilience needs to be built inside the database, inside the data structure, which means when customers are investing on our product, they don't necessarily have to go out and invest whole bunch of other technologies. Resilience, high availability, performance, security, they all are baked inside the database. So as part of the transition from Oracle Database 19C to Oracle AI Database 26AI, we have introduced a whole bunch of optimizations that drive availability end-to-end, that drive resilience end-to-end, that has built in security. And our principle has always been when customers upgrade to Oracle AI Database 26AI, they can get all these benefits by default, no application changes. That has been our design principle from day one.
Dave Vellante
>> So I also read that failover times, obviously dropping, I think platinum is under 20 seconds, then the diamond is near instant. I mean, under three seconds, zero to three seconds, I think is what I read.
Ashish Ray
>> Yes.
Dave Vellante
>> So I'm interested in a couple things, like what architectural changes you guys had to implement to enable these gains, especially without application changes. What does that mean? I mean, the application is now sort of fenced off from that. Is it any applications that's supported by Oracle database? Is it only fusion applications? Take us through all that.
Ashish Ray
>> Got it. Absolutely. First look, let's look at the platinum tier and then we'll look at the diamond tier. Platinum tier is the one where we are introducing all these changes, all these innovations that don't require application changes. By that, what do I mean? When you look at say a failover time, and you mentioned failover time now reduced to 20 seconds, 30 seconds, these are really complex applications, complex applications over multi-node clusters. What we have done, the fundamental technology that enables this failover is RAC and Data Guard. What we have done, say when you pick up Data Guard, we said, "Okay, when Data Guard does failover, there are three parts to this failover process, the redo generation, redo transport, and redo apply." By redo, I means the block, the block of the Oracle change. So we looked into each of these layers and we said, "Okay, how much optimization we can drive farther?" So here we are really looking at kernel code and driving optimization that such that bringing in more parallel efficiencies, removing extra weights, making sure things can happen in parallel. When we do that, fundamentally these changes, innovations are happening at the database layer, which means the upper level, the mid-tier, the application, they can automatically take advantage of it. And that's part of the reason customers are choosing to upgrade to Oracle AI Database 26AI. But mind you, there are two other components to this. Number one is Exadata. Exadata is our premier platform, extremely high performance, highly available, highly scalable, reliable platform that many customers are banking the mission-critical workloads on. So what we have done with Exadata plus the database optimizations, the hardware and the software optimizations work in tandem to the extent the whole is greater than the sum of the parts. And number two, when you deploy it in our cloud or in multi-cloud, you get best of all worlds. You get the Exadata efficiencies, you get the database optimizations, plus you get the cloud efficiencies.
Dave Vellante
>> Interesting. So I remember it was probably mid last decade, you guys announced your strategy. It was same-same, Exadata, extreme integration between hardware and software. You're really the first with that same-same, but the fact that you've been able to bring it across clouds is quite remarkable.
Ashish Ray
>> Absolutely unique also, Dave.
Dave Vellante
>> People say, "Oh, Oracle was late to cloud." They joke about Larry's tirade on the cloud, which was awesome by the way. But now you've got the most comprehensive multi-cloud strategy of anybody, including on prem. And it's because you're putting the sort of identical infrastructure across the globe. I want to go back to actually something you said. So you sort of took us through a scenario where you're basically simplifying that kind of recovery almost instantaneously without having to slog through redo logs and guys in lab coats having to squint through all this data, right?
Ashish Ray
>> Yes, yes, absolutely. Because this is all kernelized in some sense, all the optimizations that we are happening happens really a part of the Oracle kernel processes. There is no lab coat required in Apple Cloud. Absolutely not.
Dave Vellante
>> So historically, when you think about this level of availability, I've got to balance availability or make trade-offs, be availability, cost and complexity. I need skillsets. So what have you done to reduce that complexity? Do I still need a lot of specialized in house expertise to get to this level? What are my sort of prerequisites?
Ashish Ray
>> Yes, yes. Excellent question, Dave. And that is our second principle in some sense. When we are introducing these optimizations, we are saying, "Hey, can the database by default make certain decisions?" And you say, "Ashish, how are you making these decisions?" Remember, we have been in this space since decades, decades of enterprise mission-critical learning experience. What's happened over the last couple of decades, we are bringing all this knowledge and coming up with the most optimal setting, which means when customers deploy these databases by default, they're getting all these high performances, high availability, and we make sure that customers do not need, again, any lab coats to tweak and change parameters and tune and pray things happen. Things should happen by default, and that's what we are bringing to market.
Dave Vellante
>> I want to ask you a question and have you clear up what may be a misconception for people, because everybody's familiar with AI. They're using Chat, they're using other coding agents, et cetera. But we hear about how agentic AI is, you've got to have agentic, you've got to have a core enabler for success. But you guys are emphasizing high availability in this agentic world. A lot of people might think, "Eh, what's the big deal? If it hallucinates, I mean, I'm just chatting with the system," but talk about why it's important for mission-critical applications.
Ashish Ray
>> Yes. Yes, absolutely. All these mission-critical systems, they're already under tax, tax in the sense that there are mission-critical applications, leveraging the OS, leveraging the network, leveraging memory, leveraging flash, the hard drives. They're already heavily utilized. Now on top of this, what happens? Comes agentic AI workloads. Now can you imagine if there are layers of bottlenecks slowly developing, everyone suffers because these agentic AI workloads are making quick decisions regarding, "Okay, let me read some data really fast in microseconds and make some decisions, drive automation, drive workflows." So the essential paradigm of availability, scalability, performance, now the requirements are even multiplied when agentic AI workloads are hitting the same infrastructure or even a replicated infrastructure. That is why my advice to all C-level people involved in this, make sure you have a high performance, highly scalable, highly reliable data infrastructure that is guiding your agentic workloads. You cannot just do one and hope things work. They will not. You need a combination of both worlds.
Dave Vellante
>> So I've been an Oracle fanboy for quite some time because I kind of grew up in that era where mission-critical was super important. And of course, just watching Larry operate, I mean, I remember the database wars, it was side based, there was Informix, IBM was in the mix and it wasn't clear at the time who was going to win. And it's clear now who won.
Ashish Ray
>> Thank you.
Dave Vellante
>> And a big part of that is because it's just the intense focus on R&D and customer emphasis, et cetera. So the reason I tell you this story is because I was talking to somebody the day and I was kind of being effusive about Oracle. I'm like, "I'm really impressed with what this company's doing." And I mentioned Exadata, they go, "Exadata? Is that even around anymore?" And so I laughed out loud. I'm like, "Every bank runs Exadata." I mean, you can't walk around this area without bumping into an Exadata, but so you mean thousands and thousands of customers use Exadata and it's kind of fundamental for everything that we've been talking about, same-same. You've got multiple clouds, you've got on prem, we call it a supercloud. You guys don't use that term, but essentially that's what it's become. But I'm interested in maximum availability architecture, MAA, you call it. What role does Exadata play in that?
Ashish Ray
>> Yeah. Exadata plays a fundamental role. And I mentioned Exadata is essential for the platinum tier because with all the performance advantages that Exadata brings, add on top of that, the database optimizations that we have built, combination is what drives the enormous performance gains we have seen in platinum, the 22nd failover for complex apps. Even for applications that are designed for extreme availability and can take advantage of diamond tier with a distributed framework, Exadata is essential because it is providing the underlying reliable fault-tolerant network, HA built-in, plus the performance, again, combination of hardware, software engineered deep inside, whole is greater than the sum of the parts.
Dave Vellante
>> I was talking earlier, I think it was with Juan when he was here. I was saying two years ago at Oracle, at the time it was called Oracle Cloud World, Larry declared that the era of multi-cloud has arrived and he brought Matt Carmen out in the stage. I joked that a lot of jaws dropped. And it's funny because I remember a time when I said a company like Amazon is never going to let real application clusters inside their cloud. And maybe, I could see Azure and Google maybe with the Thomas Connect, but anyway, it's there now. You've got this not only vision, but you're implementing and operationalizing multi-cloud. To what extent are these capabilities truly same-same, and extended to AWS, Azure, and Google Cloud environments? And of course, OTI class.
Ashish Ray
>> Yes, yes. It is absolutely same. And that, again, has been our design principle. When we started the cloud journey, given our decades of enterprise experience on premises, we made sure that, hey, we are not going to fork off a new kernel for cloud. We are going to take the same kernels, same architecture, bring in the troves of knowledge that we have acquired over the last couple of decades and apply them to cloud. Hybrid cloud, that's where cloud at customer comes in, cloud as in OCI, and then multi-cloud. And when customer sees that, you see at the end of the day, when you look at large complex enterprises, no one is 100% on prem, no one is 100% one cloud, right? There is multi-cloud mixed in. There is some on prem mixed in, some cloud mixed in. And when we tell them that, "Hey, it's fundamentally the same performance, same scale, same reliability as you make this journey, make this transition," that's so much of a relief to every C-level person I have talked. Because they can see that, "Okay, I can start my data modernization journey without any disruption, without any compromises," because they get the predictable high performance as they make this journey. They don't have to take the bandaid up and complicate and integrate and hire the lab coats. No, it is absolutely the same consistent, predictable SLAs as you make this journey.
Dave Vellante
>> I remember when you guys announced the Zero Data Loss Recovery Appliance, we all laughed at the name. We said, "Okay, well, that's what it is." But I remember the time thinking, and I wrote about it, I said, "There's really no such thing as zero data loss, but you can get close." But it struck me at the time that that was at least sort of a north star that Oracle has. And now you've extended that, which is increasingly important in this AI era. So I wonder if you could talk about that, how you're making that achievable beyond just sort of an appliance into the entire infrastructure. We could distribute a database, we have multi-cloud. I mean, that concept is now becoming sort of an umbrella, not only message, but actually capability across your customers' estates.
Ashish Ray
>> Absolutely. So I will give you a two phased answer. The first phase I will answer about the Zero Data Loss Recovery Appliance, and the second phase I will answer how we have extended the concept of zero data loss throughout of our maximum availability architecture. See, when we launch Zero Data Loss Recovery Appliance, fundamentally it's a differentiated data protection appliance in the sense that traditional backups would do, "Hey, it's Sunday night. Let's do my weekly full backups." Now can you imagine 40 terabytes, 100 terabytes every Sunday night, like plundering through the network?
Dave Vellante
>> Hope it works.
Ashish Ray
>> Hope it works. And then Monday, Tuesday, Wednesday, daily incremental. No, it's not very efficient. So what we did, two things. Number one, as part of the ZDLRA product, Zero Data Loss Recovery Appliance product, said, "Hey, no more full backups. Day one, one full backup, and day two, day three, day 100, only incremental backups." And inside the ZDLI product, we have the necessary engineering in place so that we can track the change blocks very efficiently. So that's number one, incremental forever. Whether zero come in, we said, "Hey, this is an Oracle database integrated product." Remember earlier I talked about the change vectors, the redo blocks? It can also receive the redo and the redo represents the latest changes. So in the traditional backup, if you were to restore, you would lose all data since the last backup. Not the case here because it has all the redo. So hence the Zero Data Loss Recovery Appliance started off on premises, available now in hybrid cloud, public cloud, OCI, and multi-cloud. The second phase is what are we doing with our integrated products like Data Guard? What are we doing with zero data loss? Essential element for zero data loss is synchronous transport. And as you know, Dave, we have been in the industry for a while, synchronous transport is gated by latency considerations. So we looked at our network transport and we said, "Given a fixed latency, how can we make this even more efficient?" And all of that is rolled into our platinum offering. It goes back to how can we drive more efficiencies, more optimizations inside the database kernel so that the applications can automatically take advantage of this?
Dave Vellante
>> Yeah. So I mean, I was saying before, there's no such thing as really zero data, but you've basically squashed the RPO, which is essentially how much data you lose down to as low as possible. The things that have to happen for you to lose data are catastrophic. I mean, you've got to have multiple data centers get attacked by nuclear bomb or something. I mean, it's got to be something that is just inconceivable, possible, but that's what you've done. You've squashed that RPO down to virtually zero.
Ashish Ray
>> Yeah. Even in the catastrophic cases, suppose an entire data center, entire region goes down, assuming you are doing a synchronous transport of all your changes to a nearby data center... And mind you, we have done a lot of optimizations in this space so we can really stretch out the data center, suppose unfortunately earthquake hits or a power outage hits. You can automatically fail over, goes back to your 20 seconds, literally, without any data loss because you have sent the data over and you have not committed the changes till you get an acknowledgement back. And the way we can do this, we can write into the recipient's memory, hence you can get the acknowledgement back very fast. It's part of what we call Fast Sync, Data Guard Fast Sync, and hence really, truly zero data loss despite any failures here.
Dave Vellante
>> Right. So you're at a synchronous distance and the probability of both going out is very, very, very, very low, and that's how you're able to achieve and claim that zero data loss. But again, you're bringing it from not only just the appliance, that concept across the entire data state.
Ashish Ray
>> Absolutely.
Dave Vellante
>> I want to give you the last word before we wrap.
Ashish Ray
>> Yeah. One thing that really differentiates us, we've talked through a lot of technologies here. We talk through platinum, we talked through diamond, the distributed framework, and we talked through Exadata and Cloud. Dave, Oracle is really unique in this space. No, you personally have been in this space for several years. The way we are charging ahead with the innovations, with Exadata, with cloud, and with this thought that, okay, there are new forces coming in with agentic AI, with cybersecurity, with distributed resilience, and still baking this in, immunity is built inside the data, we really are unique in this space, Dave.
Dave Vellante
>> Well, I appreciate you helping me wrap up today, Ashish, and helping us in this deep dive. So thank you very much for coming into our studio here at the NYSE. Okay. Keep it right there. More Oracle data deep dive from NYSE, our NYSE Wired Studios. You're watching theCUBE. I'm Dave Vellante. We'll be right back right after this short break.
Dave Vellante