navarre.ai

ClarisTalk AI · the FileMaker AI podcast

EngageU Live: Talk about database hosting with Joey Grimaldi from ODI Tech

November 25, 2024 · 13 min · Matt Navarre and Cris Ippolite

ODI Tech is the oldest of the FileMaker hosting companies, and has been hosting databases since the 90s. Matt and Joey talk about hardware, AWS, SSD tech, AI Hosting, and more.

Apple PodcastsSpotifyYouTubeAll episodes

Transcript

Transcribed by machine (whisper, large model), 2,604 words. Names and product terms are sometimes misheard; the audio is the record.

Claris Talk AI How does AI come into play in your life and what is the impact it's going to have on hosting? Sure. So to start off with the hosting company, how we came about, we started in 1998. We were one of the first hosting companies on-scene. It was us and basically Digital Forest, who's no longer around. As far as the hosting company goes, we've gone through so many things. We've done so many changes. Following FileMaker, that's FileMaker's zone and so forth. And it's so interesting having the Linux cycle come full circle. Oh yeah, you were talking about FileMaker 5.5 Linux? Yes, yes. I don't know, I think we were one of the only licenses.

When did they, just to refresh my memory, switch back to Linux for server? What version? Technically 18, but it really wasn't, in my opinion, it wasn't roughly 18. Okay. 19. Okay. So a few years ago now. Yeah. Okay. Yeah. I couldn't switch fast enough from Windows. Yeah. Don't get me wrong. We obviously hosted in Windows because of the enterprise levels of hardware. Yeah. But for managing the Linux software, being comfortable in the Linux environment. Yeah. Yeah. Reliability, speed, footprint. All of the above. All of the above. Yeah, yeah, yeah. I didn't switch to it because I wasn't as comfortable with Linux as I am with Windows. Yeah. And I don't like Windows, but it's much more desktop-y than the Mac.

Sure. Not than the Mac. It's desktop-y. It's just older. Somewhat like the Mac. Yeah. So I was able to learn it and get by. Yeah. And you have a ton of clients. Mm-hmm. Your hosting plans start pretty reasonably. Yeah. $5 a month, I heard. Yeah. I wish. In fact, we were just having this conversation. Yeah. We used to do shared hosting. Yeah. For the smaller companies, which I really am sad that that went away. Yeah. Yeah. But now. Yeah. Yeah. It went away. Yeah. With the multi-tenancy. But I understand it. So now everybody is getting on a dedicated server. They start at $69 a month. Okay. Multi-tenancy. Let's not let that one go. Yes. It came up yesterday in the keynote. I know.

I don't really know what that's going to mean, but the word multi-tenancy was used again. I know. I think it was talking about some fairly far future thing with the Linux server. But we'll see. They were talking, in terms of multi-tenancy, they were talking in terms of scalability. Yes. Yes. And how to have. How to have one server serving many clients to have the multi-tenancy in order to have the scalability. Yes. That's really where that went in with Studio. I don't know if it's like, yeah, Studio, probably. Connect. Studio. Yeah. Yeah. Yeah. So not like Docker and FileMaker server or something like that. Nope. Nope. That's totally different. So you guys do your own hardware, right?

You don't have. Correct. You don't resell AWS like a lot of other hosting companies. No. Yeah. And that's one of the things that kind of differentiates us from other hosting companies. Yeah. Yeah. And that's one of the things that kind of differentiates us from other hosting companies is that we maintain and support all of our own hardware. The hardware is actually purpose built for FileMaker. What kind of machines? I'm guessing probably not big GPUs. No. Or CPUs? Okay. No. No. So a lot of, we do virtualization. The virtualization that we stand up, we do dedicated services. So all of this, the RAM and of course the hard drive and everything is dedicated towards that particular instance.

And then we have the RAM. And then we have the RAM. And then we have the RAM. Yeah. And then we have the RAM. And then we have the RAM. And then we have the RAM. Yeah. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM.

And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. Yeah. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM.

And then we have the RAM. And then we have the RAM. Yeah. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM. And then we have the RAM.

So we have the RAM. And then we have the RAM. No, it's very similar. I think it's much faster as far as performance. And for databases, obviously, that's really where the bottleneck is. The bottleneck is in the disk. So the NVMe makes a huge difference. We can also do block storage now. So if we wanted to take an NVMe drive and attach it to an instance, we can do that. And so for some people like to have the OS and the data partitions also created. Yeah, it's less of a big deal with Linux. I think it's a best practice for Windows, but yeah. We can also do virtual private clouds now for people. So we do that with doing VPNs, for instance, for HIPAA compliancy for hospitals.

Oh, sure. So we'll stand up a side-to-side VPN for those people so that they can access their phone, like their database, and other resources if they need to do, like remote desktop, we can do. That was actually my question. Like Citrix or remote desktop type. That's mainly on a case-by-case basis. So we don't do that. We don't kind of advertise that, I guess. It's a real specific type of a FileMaker thing, right? Like a really big application that's mature, but it would take a fortune to retool the application for a win. Exactly. And the performance. Yeah, yeah. And it's just much easier to, yeah, yeah. So AI. In fact, Cindy White, I don't know if you know the name.

Yeah, clients. We just did that up for them. And not even... It's kind of interesting. This is something that we don't host, per se, but they're Sage accounting software. We put that in the cloud, in their own private cloud, along with their FileMaker server, and did the site-to-site connection with their own private cloud. So now that we can integrate FileMaker on the same network as the Sage accounting software. Yeah. Yeah. So, like for a FileMaker developer developing AI with API calls, that might not really have much of an impact. You know, it's sort of at a different level than it was testing, I think. But what are you getting from customers about it?

What's your clients doing? As far as doing things with API calls? Well, no. Like, what are the requests you're getting for AI? Oh, sorry. AI. As I said, API. And then API calls to AI. You know, the FileMaker community, I'm really glad that this is happening. Because, like I always say to you, I come from this from a totally different background. So, from a computer science background, having programmed in Objective-C and others, coming into FileMaker. Fortran, Pascal, all the modern ones. All the really modern ones. COBOL. Yeah. Having, you know, coming into this from that perspective, I tend to be an outsider, so to speak, with FileMaker. In fact, I was just speaking with another person who was here about wanting to do a session specifically.

On, you know, how everything is not a nail. Right? You know, when you only have a hammer, everything looks like a nail. Yeah, exactly. And sometimes, as FileMaker developers, if that's all we know, we kind of get trapped into that. Totally. Well, we want to make everything in FileMaker. Weirdly, I think that's one of the exciting things about it, yeah. Exactly. Because the cases for it are so attractive to do, and also so easy to implement. Right. And that's really where... That's where FileMaker shines. Because FileMaker is really a Swiss Army now. And I think this AI push is really getting the developer community to think outside of FileMaker and start looking at, you know, resources outside of itself.

Yeah. And that obviously happened with being able to make API and policy from the script step. You know, that's where all that started. But I think the AI is really pushing the developers in that direction, you know what I'm saying? Which I think is fantastic. So, I guess one thing that occurred to me from seeing the stuff I've seen here... Is using, so rather than using like open AI for semantic searching and things like that, to spin up your own model on like a local server, which kind of is kind of obvious. So, if you have a FileMaker server, you could do a VPC kind of thing. Yeah. But that's a total GPU type, very, very different type of a server. Are you getting requests like that?

Not so much now. I think after, you know, we've engaged and this now engaged you, being so focused on AI, I think we're going to start seeing this. Yeah, I think so, too. Yeah, yeah. I hope so, yeah. Yeah. There, you know, and they're all, again, with looking outside of FileMaker, there are a lot of services out there that you can stand up, you can stand up these models in a matter of minutes. Pinecone and things like that? Yeah, yeah, yeah. After the databases? Yep. And you can, you can choose and a lot of, a lot of it, in fact, is, is, is freely available and you just have to stand up a server to make calls to. Right. And do it. So, but like a lot of people have privacy issues.

Or other issues with using, like, OpenAI for stuff like that, or Pinecone, because you don't have control over everything. But if you have your server, like you, I, like, you know, I get, I get an open source model and I need a place to host it. You can't host it on your, your local computer. It has to be, like, where you're from and where the server is, you know. Right. That's why I was thinking it would be kind of a natural offering. It could, it could be, but there's, there are things that are, that are out there, I'm thinking, oh, what is it? Oh, the hug. Hugging face? Hugging face, yeah. So are, are there, I guess I don't know, I should know more about it.

Yeah. But are there places you can actually host that where there is no risk of anybody else having access to your data? Oh, yeah. Absolutely. Okay. You can, you can choose models. They have a whole library of models that's all community-based. Or you can make your own. You can supply it with a data cell. You can make your own models. You can take models and, and if they. But you still have to have a server to host it on. That's what I'm talking about. No, but they do. Okay, okay. You can do it right directly from that interface. You can basically say, build me a server. It'll build me a server within, like, two months. Okay. And then it'll also give you the API.

Got it, got it. It'll also give you the, the URL to make the API calls. And then from FileMaker you just go do it. Help me out with some of that. Yeah. I want to play with some of that. I will. It's, it's, it's actually a lot of fun. And with, with some of those in the community, you can basically search. There's, there's like over, I think there's over, I think there's over a billion now. Live language models, there. So you can search all of these for whatever subject you want. And then you can find the ones, and they're all rated, too. Yeah. They're community-based. So you can look for one that you want, and then you can take that, and a lot of them will, will give you the data set that they use to make the model.

Oh, wow. Yeah. Okay. And so you can, it'll give you a CSV file, and then you can look at that data, or if it's images, they'll give you images. You can look at that data set, and then you can, you can add to that data set yourself, and then have it learn that new data set. So you're basically modifying. Yeah, yeah, yeah, yeah, yeah. So there's, I think there's a few different approaches to that. Like, basically, what is the source of truth? Right, right. Chris and I talk about a lot on the podcast. Because you can't trust the model to be the source of truth. The model is the logic engine to get from the prompt to the truth. Correct. And keep in mind that you also have, within that framework, you also have agents that you can bring in other ones, as well.

So I think it's a really good way to do this, because you can take it, you can use it as a starting point, and then you can add your own data into it to basically refine those answers to make sure that you're getting it correct. Yeah, yeah. It's a brave new world, man. I'm loving this. I can't wait. Thanks, Jerry. Yeah, thank you.