59:15 Webinar
Actionable Assessments: Prepare Your Infrastructure for Oracle 26ai
Join us as we explore how to analyze peak workload concurrency, optimize storage redundancy, and map advanced feature dependencies.
This webinar first aired on August 20, 2026
The first 5 minute(s) of our recorded Webinars are open; however, if you are enjoying them, we’ll ask for a little information to finish watching.
Click to View Transcript
Hello everybody, and welcome to another Tech Talk on a Thursday. I'm your host, Jason Langer, where we try to bring education and entertainment together for what I like to call edutainment. And in for today's session, we've chosen Prepare Your Infrastructure for Oracle 26c AI. Obviously, doing updates for any sort of software, e- especially databases, can always
be a little tricky. There's always a little planning that goes along. And I think over the years, as I've kind of been in the field for a while, like there used to be what I called the art and the science of right sizing, especially early on when I come from the idea of single core, single socket computers, right?
You didn't have a single core or a single socket that had 36 cores on it or whatever, so you needed to get the right sizing for your applications. Kind of got away from that over the years. You know? CPU and memory and all this stuff was abundant. You had so much of it, and it was easy to get.
But as we all know, with the AI bubble, getting memory is hard to do these days. Lead times on CPU and processors are longer as well. So if you're looking at doing that upgrade, there's some tricks and tips that you might need to be thinking about. And I've got my good friend here, Thomas Stutzman, to help us cover that.
Hey, Tom. Why don't you do a quick introduction and then we're gonna go into our first segment of why this matters now. Let everybody know kind of who you are. Hey, Jason. Wow, that's a good picture of me.
I look like I'm a lounge singer on a, a cruise boat. First off, I'm Tom Stutzman, and I have been actually at Pure for seven years. Previous to that, I started my IT career in 1978. Wow. Loved computers, loved technology, and, then I decided, I liked airplanes better.
Went and became pilot and, flew for the government. And then, got back into technology and spent, since 1991, doing Oracle things, either working for Oracle or as a consultant doing, Oracle things out there in the field, and came to Pure. And, so I've had the whole scope of technology, working as an end user, as a customer, as an engineer, and, I'm really excited about what's
going on with, the industry today because a lot of the skills I learned, well, they're coming back. Right. Right. Well, before we get into it, Thomas, thank you for that. I also want to say we've got two other folks joining us today, but they're in the background.
We've got Lester and Graham, so I would be remiss if I don't mention, folks, we want to keep this as interactive as well before Tom and I really start getting into things. Bl- use the chat window, use the Q&A. If something, if something Thomas says piques your interest, you know, or you want some more clarification, drop a chat. Or if, you know, maybe you disagree with us,
which is entirely possible as well, you know, make sure you drop it in there. We're definitely gonna be actively moderating that, and so on and so forth. But let's transition to why this matters now, Thomas, for our first segment and figure out what's going on. Okay. The, thing to drive just disappeared.
There we go. There you go. Thank you very much, Jason. Yeah. I got it back, right? We're all working fine. You're good. You're good. So the first slide that we have up here, and we're This is not gonna be a, a, how would you say, a, a, a
fire hose of slides today. We're really trying to make this an interactive conversation. The world is changing, but the answers are from the past. I know that sounds like a back to the future type thing, but the reality is a lot of things that we used to do are becoming increasingly important with this AI tax that's kind of hit
us all. And in addition to that, there's a lot of new technologies that are coming out that we really have to look at differently than we have in the past. And as Jason pointed out, that's what this really presentation is all coming together to be.
So let's talk a little bit about these type of things a- and what they actually mean to us. The key thing here is you don't need a new playbook. Much of the things that you really need to look at today so you size things correctly, you get the performance correctly, you can move into AI correctly, have already been used and have been done before. We just got away with it, for a long time
without sizing our databases and sizing our, our infrastructure. There's a number of algorithms that people have used. We'll show you what those are. But the reality is we kind of threw much of this to the side because we had such abundance of everything that was out there.
There was lots of CPU power. Oh, you know what? Let's build it in the cloud, and we don't have to worry about that. We'll get what we need when we need it. But that's all changed. In fact, it, it really is kind of upside down a little bit now from the way it was, the last few year- last 10 years or so.
And because of that, and because moving into AI requires a different thought process in much insofar as sizing, we really need to kind of go back to the way we did things previously, maybe 20 years ago, to really get an idea how we can size things in the future. So back to the f- back to the past, to the future. I don't know. It sounds like a movie title.
But, uh- As a, as a movie buff, I l- I, I'm here, I welcome it, and I'm here for it. But yeah. I, I was thinking I'm Marty McFly. I'm Marty McFly, right? Yeah. No, it's funny you say that, Thomas, because you talk about the redundancies.
Like, I can remember early on in my IT career, you know, I worked for, a mouse multimedia company, and we ran a sports website that did, like, fantasy football drafts, and, like, we were always needing more database servers and stuff like that, but we, you know, we just couldn't keep going out and buying stuff. So, like, we had to tune things.
But then as the years gone on, we would just, like- The question would be like, "Hey, we just, we're, we think we're gonna grow this much this year. Just go buy more servers." You know what I mean? Like it w- it was just go buy more hardware was the answer. It was never like, "Hey, do I tweak this setting or do I do this or that?" So and it's
like we- back to your, y- back to your back to the future comment, it's funny because, like, you said in that middle, abundant of resources, but now the tables have turned again where it's like, okay, you, you just can't go out and get all these things. Oh, a- a- absolutely. And, and you know, I br- I was brought up on Moore's Law.
Oh, it's gonna get faster and it's gonna get cheaper. Mm-hmm. Guess what? That's not happening right now. You can't even get some of the things you need. So lead times on some servers, everything has just changed dynamically. And it's hard sometimes to grasp that, but you know what?
In the 1980s and 1990s, this is how we lived. We had to size things. We, we, we didn't know what things were going to be, so we had to size things, and we came up with ways of doing it. And frankly, that kind of art of sizing has been lost.
The lost art of sizing. But between 2005 and 2025, 2022 here, the Really it was abundance. We had everything was there. We had new NAND, new faster technology. We had the cloud. We had all this thing going
on, and it was exciting. And one of the things that I even found myself doing is creating bad code. I can put it on a faster system, and it's gonna be great. And I look back at it, and I, I actually had looked at something I wrote about 15 years ago, and I said, "Oh my God, this is awful." It, like, looped over and over and over and over
and over and over and over again. And it was just You could see the processes being set up there in Linux, and you're like, "Oh my God, what's going on here?" had to be more efficient. And, and, and those kind of thought processes we have to do. And today we have this whole thing of scarcity.
The resources that we need, we might not be able to get. So a lot of this presentation is really focused on how to overcome these things. Go back to some of the sizing principles that were out there before and, and really take a look at them and understand them. And if you look at the one thing that drives this, this one slide, which is And I'll tell
you, Jason, this thing blew my mind when I thought about it because I used to do this myself. Size for the peak. Yeah. Look for the top, and then give me 50% more. And I'm like, you can't do that today. Yeah. Can you imagine going into your IT department
saying, "Yeah, I know everything's on, you know, scarcity, and things are pricing and all this other stuff. I need 50% more than I think I'm gonna need, and even though that's peak and I might never get to peak." Right. Doesn't work, right? Or, or even, T- Thomas, the, some things like I used to always say, like My boss would be like, "Well, how much do you think you're
gonna need in two years around?" I'm like, "Uh, I don't know." Like, you're making the buy now, but you're wanting me to make it last for two to three year, right? So it's, you know, you're like, you're just guessing, right? You're just throwing darts at a wall for the most part, so. We, we don't guess. We, in, in com- in databases, we have a thing
called heuristics. Right. Which, which technically is educated guessing, but we don't guess. Yes. Right. Not a guess. I've, I've met my fair share of database people. You're, you're correct. I used to love that one.
So educated guessing is okay, but guessing is not. Um- Right well, here's Jason, this is kind of crucial here because 26AI, there's a lot of other things that really are coming to bear, and we're gonna talk about some of those things and what they mean to your sizing 'cause you can't just size it as a regular database. Mm-hmm.
And when you look at, how this is all gonna come together, you could be in some com- capacity and performance issues right away if you don't look at it, you know, seriously. So let's type a little bit of this right-sizing infrastructure and why it matters more now today than it did years ago, and it's partially because of the criticality of pricing.
I mean, you're, you're, you're looking at CPUs that used to go up in performance every year and get cheaper, and now if you can get it, you're lucky. And then the same thing with other technologies that are in play there. We're gonna show you how you can encapsulate all this and really size things pretty correctly.
So let's take a step in and take a look inside Oracle 26ai and how Everpure can actually help. So every discussion, everybody wants to talk about it. Every, every time I get called into a customer, somebody raises their hand and say, "What about AI?" And- Right you know, right away you're like, "Okay, well, that's, that's a different discussion than what we came here.
We came here to talk about backup, but now we're talking about AI." And I just put in the chat, but we went 10 full minutes before we said AI. I think that's a record, so. Nowadays, I feel like that's a record, but keep going, Tom. That's a, that's a, that's actually a very good one.
So 26ai, there's all kinds of things we need to worry about. Vectors and embeddings, how to maintain them. Where are they stored? How are they stored? Are they NFS? Are they S3 blobs inside the database, outside the database? Lots of options that are there for you.
Yeah. And, and how do we minimize the CPU impact? I mean, that's, that's crucial because a lot of people You know, we, w- we are a data company, so we look at storage and data kind of holistically. But it affects networking and performance on the CPUs and all the other things. So you ha- you have to look at the whole ball.
And then the other thing too is, you know, how do we size for these things? How do we size for the future? As we mentioned before, we guess, right? Hopefully by the end of this presentation you'll have some educated guesses you can use, right?
So I'm gonna take a step back, and I know some of our audience are seasoned DBAs, and some of the audience might be people who, you know, they're looking at me right now with a vector, and they're thinking, "This sounds like Star Wars." the new battle cruiser or something. And vectors are a pretty simple thing to understand. It's a dimension, okay?
It's, it's a pointer. It's a pointer that helps AI really understand in the computer what is going on from a textual document to an image to an audio file to any one of those different things. And that information is embedded inside the database.
Now, they can be embedded inside the database, you know, actually inside Oracle, or they can be externally, but the point is that they're used by the database to understand the logic. Now, those things are loaded in, and they change just like everything else does, so they have performance potions and things like that you have to be careful about. But they're stored part of the database.
That's 26gI. And that gives you all these similarity searches and things so that way I can say, "I'm looking for brown houses on a street called Elm in an area that is, above the water table so it doesn't flood." Something like that. I thought you were gonna say- And yes- nightmare, but Yeah.
Yes, that's right. And living next to Disney, I'm not worried about flooding- Yeah 'cause I'm in the center of the state of Florida. But my brother lives in Fort Myers, and he has to worry about it all the time, and I, I remind him that I'm always here for hurricane protection if he wants to evacuate.
Now, this is really where it gets down. This, this is where the road hits us because there's different models of how these embeddings can take place, and each one of these dimensions, you got 384 and 512, 768, 1024, they dramatically increase your database. Now, I, I sat through a lot of AI presentations early on and, and all I ever
heard was, "It's gonna be 10X." And I asked them, "Well, where does the 10X come from?" 'Cause I can't go to my management and say, "I need 10X more storage or 10X more servers," or this or that. Well, it comes from all these vectors and all this additional data that we're gonna be storing, and how that data is stored. Now, this, this is a little model right here, but it kinda shows you some of the changes
that will be seen. We have, like, multiple ling- lingual support, like Chinese characters will take more space because- Sure there's a lot more of them than English. Yeah. But you can do all these things, Jason. I mean, it, it, it's great, but it takes space, right?
So what does this space actually mean? What does this growth rate mean? Well, we could show you, show you that the middle section there where it talks about the, the parameters and what you have, you can go from a small database to a very large database and very large footprint real quickly.
I'm gonna show you the math on that. And this isn't try to sell you more storage, because I'm gonna show you actually how to cut that back. Yeah, how to reduce it. Yeah, exactly. Because this is reducible quite well.
So 200K, this is actually a customer that we did some tests with. They built a 41 megabyte database. Now, before everybody goes, "Oh my God, that's huge," no, it's not. But it kinda shows you the sampling of what happens. And we've also done tests with multiple terabyte database too.
200 rows, 200,000 rows, 41 megabytes in size. What does it mean by the vector sizing? And this is the middle column where it says vector sizing megabytes. What happens? My database grows 19X, 1,832%, by using 768 dimensions.
Wow. Yeah. So I'm gonna- Now, I know your example's using 41 megs, but, like, you, again, to your point, you, you think about production and scale, those numbers get big really fast. Like- Oh, absolutely. Because- Jason sees numbers go up basically at this point. It, it, it's just an
escalator from here on out. Yeah. And this is really the heart of it. So when somebody says, "Yeah, it can grow 10X," blah, blah, blah, blah, this is why it's gonna grow 10X. 'Cause if you're describing that house on that street and the colors of the siding and the
colors of the roof and this, that, and the other things, the database stores that information and, and very importantly, that's how AI becomes intelligent. So you need more in some cases, but you also might just need less. So you have to understand your application and what that all So right-sizing the embedding, this is like, right-sizing this on the front end is a great idea.
The good news is you can change it. All right? You can grow it if you want to, and you can add more dimensions and things like that. But right-sizing the front end is probably the biggest savings you're gonna have off the front end of it, and we'll show you a little bit of how that actually works.
But as you can see, there's a lot going on here. These are just dimensions. They're pointers, and the more of them you get, the larger your database is gonna get. The more of them you're gonna get, potentially you're gonna get a lot more accuracy. But in some cases, do you really need that?
Mm-hmm. And, and that, that's, I mean, this just shows you, like, with a 50 terabyte database, which you could end up You know, you could be looking at 4 to 6X when you start looking at making snapshots and clones and backups and all sorts of other things. It can grow really quickly.
Now, right away, that's your call to action. I mean, oh my God, this thing's gonna grow quickly. What is Everpure gonna do about it, Jason, huh? Right. Yeah. Well, tell me, Tom. Tell me more. But- What I will say though is I like the fact
that there's actual, like, proof and math behind this, right? Because it's one thing you, you know, you talked about obviously we're a data company. It's one thing for us to say it's gonna grow 10X, right? Like, to, for us to go say. But for you to kinda say the math and, like, like, this is basically what's going to
happen- It just brings some validity to it, which, which I really appreciate, 'cause I, I like a, I like a chart and I like the math. To your point, we talked about guesswork a little bit, but actually being like, "This is why it's going to grow. We're not just telling you it's gonna grow. We know why it's gonna grow and how it's gonna grow." So we're, you know, it- I, I appreciate
you going through that because that, that helped me a lot at least, so Well, Jason, just to give you an idea, I work with a, an ink company, and I never realized how many shades of colors there are. Now, I should've probably just gone into the Sherwin-Williams, uh- I was about to say, have you ever been to a paint store? And take a look at that.
But that plethora of stuff that Sherwin-Williams has compared to the number of variations that they have for different colors on promotional literature and cans of soda pop and all sorts of other things is amazing. We're talking tens of thousands of changes that they could put into those colors, and some are patented or trademarked. Mm-hmm.
I mean, it, it's absolutely amazing how they do all that. Every one of those become a dimension. So think of that. How many times if you're gonna be querying on a color and a scope, every one of those things become a dimension in a picture that you might be using.
Yeah. You're giving me flashbacks. We just did a bathroom model, and my wife and I went to go look at white paint for the walls. Oh, God. Don't go there with white. Yeah. And I was like the, the amount of whi- like to your dimensions, the amount of white paint Yeah.
But- So I can only imagine when you get that across the color spectrum, like that's a good way to think about it, right? Like just how many different variations of something there can be. Absolutely. That's why AI becomes intelligent. Now, lots of growth going on here.
The first thing I told you was try to get the right sizing on those dimensions up on the front end. Look at the size of the dimension that you're gonna use, how many bytes. Look at how those indexes and those vectors are gonna come together, okay? And then size it there, because that front-end sizing is gonna save you a lot of pain down
the road. But here's the thing, and th- this is, this is, I This is evergreen, okay? One of the things that, that Everpure has talked about forever is evergreen. This is non-destructive upgrades. You don't have to take your database down, you don't have to take your application down to be
able to add capacity. And that's crucial, and that's important, and that's something that even though prices might be going up on NAND, and we don't have much control over that, in all frankness, the reality is I can still start with a right size, and I can grow further as I need it, and I don't have to lose time on my system outages.
I don't have forklift upgrade. I can build upon what I have. And frankly, I mean, I came to Pure, like I said, seven years ago because as a customer, this was the pain point that I had all the time. It also was I was a consultant too, and I said, "Oh, wow, I got a revenue source forever
migrating from one system to another system." And then Pure came to play, and I said, "Well, I better work for those guys because I'm not gonna be migrating stuff forever anymore, I guess." Yeah. And Tom, I wanna pause here just for a moment because, you know, depending on the, the audience, right, like we've got a lot of, I'm assuming, Oracle database cats like here who
might not be as close to the storage, right? Or as close to the system that's housing their database. So the evergreen thing I, I do just for a second wanna d- drill on is, like, this is a, for you all, is a game changer if you're talking with your storage people for all the reasons that Tom has just mentioned.
If, you know, I've done storage upgrades, I've done forklift upgrade, I've had to coordinate outages where it's like, "Hey, guys, we gotta take your database down this weekend 'cause I'm going from storage array A to B," and maybe it's within the same company, but it's like the new v- the new model came out. I went from the X to the Y or whatever.
This is something that from a Pure perspective or an Everpure perspective that you don't ever Your, your storage people don't have to think about, which is something you can be like that means you don't have to think about it. So it's not just capacity upgrades, it's controller upgrades, which means the boxes get bigger, stronger, and faster with no downtime, no outages, and no anything.
So this is a key thing. I know, I don't wanna s- like as far as your presentation, Tom, it's minor, but like I think that's a key fact where it's like for database folks that might not deal with much as the storage. Beyond what you're gonna tell us, keep going. This is another key differentiator of why Everpure, maybe you're looking at us, you know,
you know, for, for your database. Like this is another differentiating factor that might not w- swim in the waters with your typical database person, right? Like you might not ultimately see the value of that, but that's something that you could pass along to your storage peeps.
After this call, you're like, "Hey, have you looked at this? Because I want my weekends back. I don't want to have to worry about dropping tables or doing whatever when you do a storage upgrade." So little- Hey, Jason. That, this is- Little PSA over there, Tom, so This is the whole reason I work for Everpure
because I recognize my consulting business where I was moving people from new system to new system all the time, I gotta get with the game plan, and these and Everpure was able to do it. And I didn't have to forklift upgrade. I didn't have to go through migration headaches. Once I got it there, and the, the customers that I had as a consultant, they're actually
still running on the same technology, and they've never taken their databases down yet. Yeah. So it's- That operational simplicity in is key, so Absolutely. So th- this is, this is a very This is one of the core foundations that come to data. Now, here, Jason, this is also re- really key too.
Everything that goes on today, we just talked about all these vectors and all the sizing that's going on. Well, it actually compresses quite well, but you gotta do compression correctly. And that's what we're gonna discuss right now and what that actually means. Now, a lot of people, a lot of organizations are saying, "Let's We're gonna encrypt our
data, and then we're gonna go and compress it, or we're gonna do some other things with it." And I just want to set the level straight. There's lots of good technology that Oracle has, and I'm not bashing Oracle here. But think about what you're doing. If you're going to use TDE compre- and, and compression, or if you're just gonna use TDE,
understand the fact that if you wanna get the smallest size footprint, you need to compress and then encrypt. Mm-hmm. And keep that in mind because that will come back over and over and over again as far as what happens if you don't do it that way. Now, TDE inside Oracle, which is their way of doing encryption, you can do
that a variety of different ways, but you have to maintain it, which means if you have a new table that comes out that requires, a credit card numbers per se, that weren't originally part of the original customer table, you're gonna have to go out and crypt those, which means you're gonna probably have to go through some lengthy process of encrypting it and then managing it. And then let's say you're gonna add a new CBD
or CCD or whatever code that is, that you're also gonna have in that table. You gotta go back, you have to do put a part of your encryption. So there's a manual effort that goes in play here, and I don't wanna de-emphasize the fact that, yes, there's lots of automated tools out there, but it's required for you to keep track of these things. And let's face reality, not everybody stays in
the same job for 30 years. Okay ? You need to have good documentation and need somebody to manage it. So keep that in mind. First point, compress then encrypt.
You're gonna get smaller space. And why do we need smaller? Because we got a lot of vectors out there. Right. Then when you look at encrypting and then compressing or doing some things like that, you're not gonna get any savings of space, and
with the size of these databases and trying to right size, space is important. Now, that plays into what we can do. Now, first off, Everpure offers the ability, when we get data, we compress it and we encrypt it and we put it out onto the array, and that gives us the most fundamental, smallest footprint.
All the deduplication- all the things are managed, and it's the smallest footprint, and it does not require any server processes. It does not require That gives you your encryption at rest and your smallest footprint possible, and optimize the performance because the array is doing all that work that normally a database server would have.
So that's, that's crucial to keep in mind. Now, if you wanna work with Oracle Advanced Compression, which is a fine technology, I've used it for years, there are some things you need to understand. If you get a lot of workload, let's say you have a lot of processing that's going on, and you're doing an ELT process where you're moving large chunks of vectors vectors, get it,
vectors into the database, they need to be compressed, and you're usually moving a lot of this data simultaneously in a lot of pa- you know, a lot of passes of that data. Well, that requires building redo and all sorts of other things. And what Oracle does, it sort of has this deferred recompress that goes on. It says, "Don't take away from the front end.
Just put it out there, and then we'll go back and we'll recompress it again." Now, on the surface that sounds, oh, that's great. And oh, by the way, we do some of that too because we have sort of this, how would you say, DeepReduce functionality that goes on. We kind of look at stuff later on and compress it even better.
However, Oracle, when they do it, you now have to a sort of a, an amplification of redo activity and writing and reading to get that data back into the server to recompress it and then to put it out onto the storage array. So there's a lot of extra work, extra processes, extra CPU that sometimes is not accountable.
And in addition to that, it's not sending all this data in a compressed format across your network as you might have thought it would. Yeah. Some of it could be, but that means you're not getting all the optimization. When you look at the way we do it, where we encrypt it and compress it, or I'm sorry,
where we compress it and then encrypt it. Ugh, I'm looking at the slide and getting all confused. That is the most fundamental way of getting the smallest footprint and most optimum performance, and this comes out of the box with our technology day one. Yeah. No license cost.
It's there. Isn't that great? I mean, I like it . What is that Mikey? I like it too or something. No, I do have a question for you- Yes Thomas, because and hopefully, this might be a little bit of a landmine, so I apologize, but like, I look at the, the box on the left, right?
And you've got the licensing cost, which obviously sounds like an extra licensing feature, right? So right sizing can also include budget, let's be honest, right? But when you see the I see the CP load, and I'm a s- I'm a server guy, right? Like, I- my career was data center servers, and like, running apps and, you know, having
app owners. When I see CP load, like, do you have a Is there, like Is it at a depends or is that You know, is that a 10% hit? Is that a 20% hit? You know, when a You know, this is I gave you the get out of jail free card if it depends,
but like, when I see anything that's like host side CPU load, I mean, to me, I'm like, that screams, okay, this is a right This could potentially be a right sizing problem because if I'm doing 10, 15, 20% overhead on something that I can offload somewhere else, like, that's, that's real savings and that's real right sizing. Y- you brought up, Josh, probably one, one I'm sorry, Jason.
You brought up one of the best pieces that sometimes is overlooked in this slide. I congratulate you on that. Oh, thank you. And that wasn't scripted, by the way. Yeah. So here's, here's, here's something to think about. If you open up the Oracle manuals, and having
worked for that company, I, I know where these are. In fact, I think I actually have the manuals at the bottom of the slide here to quote it. You will find that they'll say, yes, there's an overhead, and it's about 2%. Okay. Well, that's, that's understandable, right? Sure. 2%. That's a rounding error.
2%'s a rounding error. Exactly. What they don't really emphasize or tell you, and it makes sense that they don't, is that when you have lots of IO coming in and you do this defer recompress later, what ends up happening is there is all this additional workload that happens down the road. So all of a sudden, a workload, that 2%, which sounds manageable, becomes a lot more because
you're doing new redos. You're, you're And redo activity requires lots of CPU process, lots of writing going on, and you're moving data back and forth. And when you do that, you're upping your CPU, you're upping your performance. And one of the ways, a- and I found this out actually a very unique way, we did some tests
with a customer. We took a look at it, and they were running backup at nights, and they were saying, "Our backup is running three times as slow as we thought it would. What is wrong?" And we're like, "Oh, well, you know, let's take a look at it." And we're looking and we're looking and we're looking. And all of a sudden, I look at it, I said, "What are all these compressed processes out
there going on?" And what it was is that it put everything in this deferred category and started running it in the evening- when they were doing their backup. And what do you think was the outcome of that? Well, backup got slower. It was doing hot backups or, you know, essentially keeping the application and the
database running. So it was contending for the same stuff- Sure while it was trying to do backup as all this movement of compressed data. So you have to understand that 2%, that's document. But in many areas, in many ways, you're taking away from a, a limited, a limited pool of resources, and it's gonna be much higher than that.
Yeah. It's gonna be much higher than that. Make sense? Makes s- It's a trust but verify statement, right? Like, it's 2%, but depending on what else you're doing, mileage may va- I mean, they're not gonna say that, but that's, that's basically what I took from you, was like this
is what they're gonna say. Run your processes, test it out, you know, there's other things that are gonna influence that, so. Well, J- Jason, I gotta tell you, I, I wanted to use my Ronald Reagan i- imitation right now, "Trust but verify," but I think that truly shows my age. It, it, it might. It might.
I g- I got it, but yeah. It's, that's on the, on the break right there. So there's other things that we do which are really, really cool, and what I've found is that a lot of organizations today are saying, "You know what? We're building this database.
We're trying it out. Once we get it down to the way we are, then we gotta go back and we gotta do it again and again and again." The sizing of the storage, the how we're gonna set up the volumes, and all this other stuff. And one of the greatest benefits that, that Everpure has come out with is Fusion. This whole idea of doing presets, it truly allows me to right-size things, see how it's
gonna work, and then be able to capture that and put that out over and over and over again. So I'm always in compliance. I'm always in compliance to the configuration, and I can move my application data and such like that around as required within the Fusion architecture. This is awesome. This is actually awesome because a lot of this
work you do and you say, "Yeah, I got it right," and then somebody comes in and says, "Well, we're gonna do this or we're gonna do that." If you have a standard and you have the ability to set up presets, then you can duplicate it very, very easily, and that's- Yeah, or they come in and say, "Do Now do that 50 times." You got it. Yeah. And do you think anybody
builds just one AI database? Ugh, no. Yeah. You know? Ask the company that has just one database out there. They're gonna be like, "Hey, that worked well there.
Let's try it here. Engineering wants to do this. Okay, sales wants to do that." All of a sudden is you have all these islands of information being created. Right. And you wanna do it efficiently, so.
Yeah, and Tom, I'm gonna pause you just for a quick second because we know what Fusion is. But again, for, for somebody that's either new to Everpure or maybe they're not up to latest and greatest, my, my, my ten-second blurb is Fusion is a, is a control plane that spans across all of your, Pure arrays. So if you've got five FlashArrays or you've got 500 or you've got FlashBlades, which is
our, you know, object and file, you've got a common API control plane where you can, to Tom's point, issue these presets and then span it out across multiple arrays. And I'm very simplifying it, but just know when we say Fusion, that's what we're talking about. It's, it's a control plane to make automation and tasks at scale, or even if, you know, maybe if it's two or four.
Like, it doesn't have to be 500. But I'm just saying nobody wants to be, you know, clicking the same mouse buttons or keyboard buttons multiple times. So just a quick level set there just in case anybody Like, you hear Fusion, you're like, "I don't know what that is," so.
It's all about making your life simpler, right? 100%, yeah. Okay, so we've gone through this tour. We figured out that, hey, the world's changed, right? And oh, by the way, AIX AIA, AI is causing all sorts of
scarcity with resources and things like that. And then we looked at some key points of what Oracle 26c AI with vectors and things. Well, now let's take a look inside Everpure Oracle Infrastructure Assessment, which is something that we've created, and it's unique. It allows us to really look at your ecosystem and help you decide and define where you need
to go. And we're gonna take a look at this. This is, this is actually crucial because what ended up happening is a lot of the intelligence that people used to have, and they did as normal practices, well, we've lost some of that. And what we've been able to do is we've been able to develop internally a number of ways to make you effective and efficient without
necessarily having to relearn the whole process. Yeah. So I, I g- I get excited about this because it makes my job so much easier. But before we get there, let's talk about the 95- 95th percentile. Ooh, exciting stuff. Yeah. This goes back, Jason, to our first thing we
talked about when we said, "Hey, you know what? We used to size for the peak, and then give me 50% more so I- I'll be able to survive and maybe till my next job." Well, can't do that today. We, we really have to be more efficient. What ends up happening is you got these two wonderful little curves here, you know, the
standard bell curve, and then you've got this P95 curve that sort of is up in the front and kind of scales down. What we're trying to do with P95 is we're trying to say 95% of the time you're gonna hit it- Yeah somewhere there or less 95% of the time. That 5% of the time, essentially, and we might call this in calculus, like, outlanders and
stuff like that, the reality is is that you don't scale that often to that point, otherwise it would be one of the 95 percentile. So 95 is kind of a good place to be, but that's not ev- for everybody, but that's one of the algorithms that, in a sizing methodology- Yeah and the one that I've employed over, 30-some years, works pretty well and is pretty efficient.
But let's show you how that comes into play in our assessments. Yeah. The p- the 95th percentile, I mean, from my consulting days, like, yeah, that was, to your point, like, well, and your peak might have been some random one-off task or whatever, right? So again, when you're buying and sizing, you know, it doesn't have to be peak or, you know,
5X above peak or whatever. So yeah, the p- the p ni It makes me think of P90X when you said it that way, but, yeah. I was- It's definitely the 95th percentile, so. I'm gonna start doing those exercises. Yeah. So what our sizing w- what, what the
assessment actually does is that it takes a look at AWR reports. But we don't look at just one. We ask you to send us, essentially what we like to call a business timeframe, so one month worth of them. Now, that's a lot, but what we do is we assemble from that a really good understanding of what you're doing, and we can help you
understand what you need to size and where you need to go, and that's important because it's the whole business part. Yeah. We had one, one organization gave us one AWR report. They said, "This is the busiest day of the year," and it was in the middle of June.
And we're like, okay, we gotta, gotta trust them. That's what they said. And, they're a financial institution that, deals with, a lot of purchasing, specifically around Christmas, and I asked them just one simple question. I said, "Well, if this represents what, what do you do at Christmas time during Black Friday?" And the room went silent.
Understand the business, and that's one of the things that we bring to the table. We ask those questions. Yeah. Is this truly representative? We can go with it. But here's what the summary, sizing summary does, and it's one of the first pieces, and there is a, a plethora of information that the
assessment actually provides. The key thing is, is that we start out pretty simple. We tell you, "Here's your peak. Here's your P95. Here's your P90, and here's your average." Please don't use average.
Average is- Yeah is normalized across everything. You wanna see that P95, and we provide you that. We tell you what that is, and we help you study that so that way we can build out your entire architecture. Now, how does this come into play for when using vectors?
Vectors are somewhat of an unknown when it comes to performance except the fact that we can actually algorithmically figure out out. You're glad you paid attention in that class. To figure out where things are going to go pretty much using heuristics as your good guesses, right? And give you some good indication.
Oracle provides a lot of this information, and that's And we've encompassed this into our assessment, so it's pretty easy to actually figure out. Now, one of the key things everybody looks at and they say, "Oh, well, we have 20 databases," or, "We have 15," or, "We have 500 databases. What do we do about that?" Well, whatever number you have, look at the ecosystem that
you're building. Figure out how many you wanna put on that ecosystem. That includes your storage, your network, and your servers. That's your ecosystem. And then we'll take a look at those individual databases, and we'll put them on this peak window analysis, which will also show you the
P95 and such, and we stack them. Now, I'm gonna show you this, th- what is really key about this slide, and Jason, this, this gets me excited when I say this. The effective peak was 41,000 IOPS, but at no time, no time did they ever come close to that during the day, any time.
That was peak on one if you added up all the peaks together. Whereas the normal processing was actually 30-plus, 36% less. So right away we were saying, "Well, you know what? Your performance, you're asking for this, but do you know you're really using that?" It's nice to have headroom, but you're not ever actually going to use it.
Yeah, and that headroom, that extra headroom costs money. Exactly. Yeah. When we had abundance, hey, just throw some more CPUs out there, which makes, you know, Oracle really happy 'cause they have lots more cores they can add. So this allows you to really tune based on Now, let's say only two of these databases you're
gonna use. Fine. You got the numbers for everything you can see. You can start stacking more onto your storage, more onto your network, because you have a better idea of what is actually going, taking place. So we started talking about, several slides ago, compression, encryption, and all these
good things. There are many organizations that I ask one question when I go in, I said, "Well, how many databases or how much of your database is encrypted or compressed?" And many times they say, "Oh, we're not using it." I'm like, "Okay." And then we run this report and it tells us they are using it. And that's actually important because sometimes we learn things about their own
architectures that they don't know. If I get a report back like this, I'm gonna say, "Here. Here's a, here's an AWUR script," which is an Oracle terminology for analyzing things, "And I'm going to look at just what you're doing with compression, how much you got, how much compression you're doing, how many tables," all that piece of it, so that way
when we size, we have a much better understanding. And if you're using encryption, we can tell that, too. So we start learning more about your environment than you might know about the environment yourself because of the study information and how it washes out to us, which I love. This is my piece of resistance, as I like to
call it. I sat there into many, many, many Oracle 26AM and 23AI previous to that, presentations, and I kept asking the question: I kept asking the question: What are the best practices? What are the best practices? And I finally got a list of about 20 things that are best practices when migrating
from 19c, which is an older version of Oracle, all the way to 23c, the, the AI, and then 20- then 26AI, which there's a long story about all that. And I wanted to know what is required, what are the things I need to look for so I can give a good analysis and good sizing. So we took those best practices that Oracle published and we put it
into the logic using AI. We used AI to make AI tool. Interesting. And it came out and it gave us, "These are the things you need to worry about." So this is based on the Oracle stuff that's out there, and we looked at everything from, you know, s-
storage requirements to network requirements, CPU requirements. Do you have a lot of bad queries out there that were written by Tom 20 years ago? And by doing that, it gives us an indication of where you are, what you need to worry about. And this is yours. I mean, once it comes out of the report, we give it to you, we say, "We can help you with
X, Y, and Z, and then here are some of the other things you need to look at, too, to get a better sizing for your environment." Pretty cool, huh? I love this. Yeah. I mean, I, again, I've As like I did consulting work, and like re- I'm very familiar, like seeing these reports and reviewing them, and it's always It's very
telling how you can show these things to a customer, and we can educate each other, right? Like, usually I got brought in, like, "We're having a problem." Like, okay, I would ask these questions. We would run a tool. Then they would ask me You know, we, it would s- it, it's, it was a joint conversation as opposed to sometimes me just going, "Here's a report," you know, "This is what your
challenges are," right? Because you really have to know, like customers should really know their environments, but they also, but they don't always know what everybody else is doing per se, right? So you can bring those best practices, and, and I say air quotes, right? Because best practices may vary on what you're doing and the types of workload.
Like there's a lot of caveats to that, right? But the ability to be, to give numbers and then have the follow-on conversation about numbers between, between teams or between folks, that's really where these kind of assessments to me really shine. And I can just tell, like, just by by the way you're presenting these slides, right, like,
you know, it, it's important. You know what I mean? As opposed you're like, "Ah, here you go. Here's a slide with a bunch of, you know, red, red are good or red are bad, green is good, you know. Go, go figure it out," you know? So, that's definitely something that, that, that, that comes to the table, right?
Like you've got all this experience that you can bring to bear and share, as opposed to other folks on our teams as well, not just Tom, by the way. But also just being able to help the customers as well, right? So it's, it's an education going both ways, so I think that's, I think that's great. That's, and I lo- I, I love these kind of reports when I was in consulting and stuff
like that, too, because it just opened those doors for better understanding and communication, like, "Let's solve the problem. You're, you're having a problem. We brought everybody together. Let's figure this out." Well, I wanna add one thing, Jason, that, on our community blog site, I actually took that report and I published what all those things actually mean and what
are the m- logical ways of correcting them and using them. So this is open. I mean, you can look at it and get the report- Yeah and, you know, a, a person from Pe- Everpure will help you understand it. But if you wanna go back and take a look at some of these things, it's, it's all published.
And I tried to make it so it was very transparent of what this is actually going on, what are the things we're looking for, and why are their importance to that, and that's on our community blog. And you're more than any Everybody is more than welcome to go there and look at it. Trying to drum up business and people looking at it, right?
Yeah. All right. Few more, few more things to talk about here. One of the key things, our FlashArray architecture, which is one of the things that we talked about today from a storage perspective, has really awesome performance. And why do you need awesome performance?
Well, you're gonna be putting a lot of workload on it. When you start having all these vectors. And by the way, vectors actually compress quite nicely. Vectors are things that when you start throwing it into, an all-flash environment like we offer, you're gonna get good performance out of it.
And then you take a look at all the other things that you can provide where we can move things into the cloud, we can build out applications and things like that that can change, and we use the Fusion to move things around. There is a number of different technologies that are really built on data, and that's the key thing. This is, this is not about throwing a bunch of
bits into a spinning drive as, as storage vendors would talk about years ago. This is actually, how do I build an ecosystem that's going to be high performance, smallest footprint possible, being able to change on a dime, being able non-disruptively so I can keep my applications and my databases and everything up and running? This is all tied together, and this is where we've built an entire ecosystem to do that,
and Oracle works great on it. I also wanna put in an awesome plug. So I've talked about our assessment. Let's say you're done with the assessment. You have the StuffStore, you know, installed.
You're running it and you're happy, and all of a sudden somebody comes with that ubiquitous database out there that they say, "Hey, you know, we, we would really like to do some 26 AI with this database. We are excited about how you've automated your systems and you're using AI inside Oracle today.
What will it do to our capacities? What will it do to our performance?" Yeah. "What will it do to our architecture?" And Pure1, which is built into Pure, okay? That's, Everpure's technology, is absolutely awesome because I can model things.
I can take a look at that AWR information, take a look at the existing performance off my array, put those things together, and be able to see what capacities and performance are going to look like. And this is You know, one of my favorite things in the world is to demonstrate this to people, and their eyes just light up, especially DBAs who are saying, "Yeah, we
think we're doing this," and then you show them what they're really doing, and they're like, "Wow, I didn't realize that was the effect," or, "That was the, the, the accumulation of our work." And this is, this is quite important to have this technology. The other aspect of the whole, you know, Pure1 technology that's out there is that it's so cool because it looks at 1,000 different pointers.
Yeah. It really is taking a look at that whole spectrum. So I'm not just looking at one individual thing saying, "Oh, this is what you really need." I'm looking at it holistically. Pretty cool, huh? Yeah, I was gonna I did wanna inter- Was
waiting The, the Pure1 stuff for, for, again, if you're new to Everpure, like, and this is, this is included. Like, the, the No better way to say it. Like, this is an add-on or whatever. It's in there. It's in there.
It's in there. It's a web-based tool. You know, you can see the performance numbers, and I really like One thing that a lot of folks don't talk about enough is the, is the capacity and the sizing, Thomas. It's like, that's a great thing. Again, I know we're talking about right-sizing the whole, you know, theory of a lot of this
is you can go in there and say, "Well, what if my workload did this?" Or, "What if I add that?" And you can see it, you know, say like, "Well, this is what the, the impact would be on your systems," or if you would need more capacity, or if you would need more system for performance. So the capacity and sizing for the Pure1 Planner can't be, overstated.
It's also a great tool for, like, performance troubleshooting, of course, but, like, that's, that's only one side of the tool, so I love it. Makes my life a lot easier using that. Yeah. So the value, you know, we wanna be, we wanna be a 26 AI value partner, and you know, it's not just 26 AI, any type of Oracle
implementation you have. I have customers today who are still even running on 12.2, which I think is, like, 15-plus years old. All that w- we can provide value with, and we can tie everything together. So we have the ability, we have the capability to give you the performance, to make things easy.
You know, I, I, I remember years ago when Windows, came out, Windows, and Bill Gates was famously said that, you know, "We're gonna make it simple. We're gonna make it easy. If it's not easy and it's not simple, people aren't going to want it." Well, we kinda had the same effect here when our founder came out and said, "We're gonna make the storage, we're
gonna make it simple and easy. We're gonna take the complexity out of it so people can manage things and work with things." And that is essentially a, a theme that goes inside our company. It's part of our DNA from early on to really figure out, let's drive out the complexity and make things easier to use.
That Pure1 exists. That's why when you create volumes, that's all you need to worry about. Everything else is handled within the store. That's why we have the whole concept of doing compression then encryption built in, making it simple and easy.
And then 26 AI brings in all these additional things that you have to potentially worry about. The reality is w- we've taken a lot of that out. We looked at the sizer capabilities we have. We looked at our assessment capabilities.
We looked at, you know, how we're gonna be able to shrink the size and do deduplication and things like that. All this tied together, really makes it a platform that you wanna think about and look at going down the road. You know, I always like to Jason, this is always funny. People always talk about, you know, the whole AI, and I kept, my first presentation seven
years ago coming here, I talked about AI and how we use AI internally. Yeah. We have experience doing this. Absolutely. So to sum up everything, maximum performance, minimized footprint, size it correctly. The art of sizing never went away.
We just stopped practicing it. But with 26g AI and some of the things that are happening today with the economics, this is the place to be, because we make it simple, we make the footprint optimized, we help you be able to assess your environment and be able to size it correctly, and then we, we unlock all that efficiency. And that's really important, is getting it
down to efficiency, getting it down to the smallest footprint, getting everything to the point where you're gonna optimize your CPU resources. And remember, the most expensive thing you can ever have in an Oracle environment is a s- a server with lots of CPUs, because that jings up a lot of bells on the cash register, telling you that you got more licensing costs.
So try to move that workload to the right place at the right time, and that's, that's one of the things that we can do. Well, that brings us to concluding, okay? Yeah, you s- you told us all this good stuff, but, like, how do folks do it? Exactly. So you can connect with an Everpure advisor.
You can get the assessment. This is something that is done, gratis. We can analyze your environment and help be able to put it together, and be able to give you that report back, and actually somebody get on the phone, or get on a webinar or web- or Zoom call and work with you through it.
And that's all important to really get an understanding of where you are and where you need to go. The other thing, too, is all that is, you can go lin- take the links that are provided, and that'll get you involved and integrated into the system so we can, you can reach out to us and take a look at what we can offer. And that's available, today.
Yeah, so if you're an existing Everpure customer, you know, hopefully you know your account team. Reach out to your account team. They'll know how to get a hold of Tom and our other database cats that can help with this. Or reach out to your, your, you know, your partner of choice. You know, that's, that's gonna be the way.
Or if you're researching and new to Everpure, reach out to your partner. Tell them, "Hey, saw this great presentation. I'd like to get more in contact with them." They'll definitely know how to reach us as well, so that's just another avenue if you don't know, like, potentially your Pure account rep or your Purity team.
So just know that you can go to your part- your favorite partner of choice, let them know, and they can know how to get in contact with these fine gentlemen as well. I know we're basically bumping up to time, Tom. I don't see any questions, so I g- I, I'm not gonna siege anything, 'cause I wanna be respectful of people's time to end on the hour mark.
But if you do have questions for Tom or, you know, Lester or Graham, we did mention our community site. These guys, troll it. That, troll's not the right word, but they're active on there. So if there's something that came up in today's session or today's presentation that you wanna know more about the assessment, or just more about Oracle, or just more about
Everpure as a whole, this is a great spot for it, beyond going to, of course, everpure.com and seeing the corporate side. This is where myself and Tom and all of our experts kinda hang out and answer questions and whatnot. So this is where Tom's blog, I believe. Is this where your blog is, Tom, that you referenced earlier?
Or- It's the community site, yep. Yep. So all that kinda stuff, go check it out there. Kind of a one-stop shop in that regard. And if you made it to the end, we're at, like, one minute before the hour, I wanna thank you for joining. Congratulations to James.
I saw that you won the raffle. That's awesome. Beyond winning the raffle, I hope everybody else ha- took something away, hope you learned something, and like I said at the beginning, I hope you were also somewhat entertained for some edutainment. But thank you again for your time.
Until next time, hope you have a great week, and Tom, thank you for coming on and presenting. I really appreciate it. Thank you, Jason. Awesome. Thanks, everybody.
Thomas Stutesman
Principal Field Solutions Architect, Everpure
Jason Langer
Senior Technical Evangelist, Everpure
As organizations adopt Oracle AI Database 26ai with capabilities like AI vector search, understanding whether their existing data infrastructure can support these new workloads becomes increasingly important, while keeping costs under control.
Join us as we explore how to analyze peak workload concurrency, optimize storage redundancy, and map advanced feature dependencies. Attendees will walk away with an actionable framework for evaluating their own environment, including access to a complimentary, privacy-first Everpure readiness assessment to help automate the planning process.
Key takeaways:
- See your true workload and capture a consolidated, hour-by-hour view of peak demand.
- Crush storage bloat and learn how real-world environments are safely shrinking their storage footprints.
- Learn how to align infrastructure for active features like TDE, HCC, and Smart Scan without over-buying hardware.
- Secure your assessment and get exact sizing recommendations.
We Also Recommend...
Personalize for Me
My Updates
No updates at this time.