325: Db2 or Not Db2: That Is the Backup Question

[00:00:00] Speaker A: Foreign. [00:00:08] Speaker B: Where the forecast is always cloudy. We talk weekly about all things aws, GCP and Azure. [00:00:14] Speaker A: We are your hosts, Justin, Jonathan, Ryan and Matthew. [00:00:18] Speaker C: Episode 325 recorded on October 7, 2025. DB or not to DB that is the backup question. Hello, Ryan, how are you today? [00:00:28] Speaker A: Why, hello. [00:00:29] Speaker C: It's really hard to do that introduction. [00:00:33] Speaker A: And now Justin's got it down to a science and then you have to segue into a thing. Yeah, yeah. [00:00:37] Speaker C: He just like magically segues. I'm like, I don't know how you do that. Well, it's just us today with attentive listeners, you'll notice. Yeah. For our five listeners today, which probably will be a Ryan and a Justin, but don't worry about them. So let's kick us off with AI is how ML makes money. OpenAI Sora 2 lets users insert themselves into AI videos with sound, introducing synchronized audio generation alongside video synthesis, matching Google VO3 and Alibaba WAN 2.5 capabilities. This positions OpenAI competitively in the multimodal and AI space with what they call their GPT 3.5 Moment for Video, the new iOS social app, because that's always the way you solve everything was with apps. Features allows users to insert themselves into AI generated video through cameos, suggesting potential applications and personalized content creation and social media integration at scale. SORA 2 demonstrates improved physical accuracy and consistent across multiple shots, eliminating all the randomness that used to happen when you would teleport or randomly become deformed. The model can now simulate complex movements like gymnastic routines. And while maintaining proper physics better than most games that you play. The development signals an increasing competition in the video synthesis market, with major cloud providers likely to integrate similar capabilities into their AI services portfolio to meet the growing demand of automated content creation tools. So now you can before where you could sort of trust social media videos, now it's all gone, Ryan. You can't trust it. Ever assume that everything is AI videos now? [00:02:23] Speaker A: I already do. Yeah. And I assume that, you know, I already thought this capability existed out there because I haven't really played around with social media a lot in a while. Like it's wherever, really, if I'm honest. [00:02:37] Speaker C: Yeah. [00:02:37] Speaker A: So, yeah, we'll see. I mean, I don't know, I think I can see how this is kind of neat, but I'm sure it's gonna also be behind a whole bunch of terrible content. [00:02:48] Speaker C: Yeah. I saw someone post like, now you can no longer trust. And it was like Benjamin Franklin with Tupac standing next to him and he's like, these do not belong in the same videos. You know, it was like him also coming out as, you know, him, like on a motorcycle. He's like, I'm terrified of motorcycles. Why would I ever do this? And you know, so you're just gonna see all these things that, you know, look realistic but now really aren't. So you really have to question. I feel like a lot more of what you see, which I'm already pretty cynical, so I already question most things I see. But, you know, now I have an excuse just to be cynical, which before I was just a crazy person. [00:03:24] Speaker A: You're still crazy. It's just, you know, the world's catching up to you. So. [00:03:27] Speaker C: Yeah, I have a reason to be crazy now, Ryan. [00:03:29] Speaker A: That's right. All right, well, moving on. Joules introduces new tools and APIs for developers Joules, Google's AI coding agent that shouldn't exist because there's already Gemini Code, now offers command line access through Joule Tools, an API for direct integration into developer workflows. Moving beyond its original chat interface to enable programmatic task automation, joules API allows developers trigger coding tasks from external systems like Slack CIC Pipelines and enables automated code generation, bug fixes and test running as part of existing development processes. The recent updates include file specific context selection, persistent memory for user preferences, and structured environment variable management, addressing reliability issues that previously limited its production use, which no and hopefully what these updates position Jules as a workflow automation tool rather than just a coding assistant competing with GitHub, Copilot and Amazon CodeWhisperer by focusing on asynchronous task execution rather than inline code completion. The shift to API based access enables enterprises to integrate AI coding assistance into their existing tool chains without requiring developers to switch contexts or adopt new interface. [00:04:46] Speaker C: To feel like we're just adding to the tools, then you got to figure out which one's actually going to be useful for you because so many of these I feel like are close in feature parody with Claude code and now Jules code and everything else going on. [00:05:01] Speaker A: It'S just Gemini code and codewhisper. Yeah, now there's a codepilot. [00:05:06] Speaker C: Also, my wife's name is Jules, so every time I hear this I'm just like no, it's got to be very confusing. Yeah, no, no, it's just she just says why are you in the black box again? OpenAI doubles down on chip diversity with AMD OpenAI signed a multi year deal with AMD for chip processing requiring requiring up to 6 gigawatts of power plus an option to acquire 10 tens of billions in AWS in AMD stock, diversifying its heavy reliance on Nvidia GPUs access through Microsoft Azure. AMD's partnership joins recent deals in including 10 gigawatts of Nvidia GPUs with 100 billion in investment, a Broadcom partnership for custom AI chips in 2025 and 300 billion in Oracle Compute deal. The diversification could benefit the broader AI ecosystem with increasing competition in the AI chip market. In theory, it should helpfully lower price and reduce supply chain vulnerabilities from everything that's going on in the world. AMD expects tens of billions of revenue from the deal because why else would they have made the deal marking a significant valuation of AI technology? I mean, I feel like this is going to be good for the end consumer. I mean so many people can't, you know, if you're looking for anything personal, you know, the running joke is, you know, can Jonathan just buy one GPU and we'll just all use it for our own stuff. But you can't even get those GPUs. So hopefully this will help kind of clear out the market a little bit and make things be a little bit easier so that, you know, more companies can innovate. Because right now everyone's stuck on just the Nvidia GPUs. [00:06:52] Speaker A: I'm stuck on this article sort of using gigawatts of power as a unit of measurement for gpu. Like that's hilarious to me. Like we're just. [00:07:03] Speaker C: I. [00:07:03] Speaker A: There's not this many, not this many GPUs, but like this much in power of GPUs. [00:07:10] Speaker C: My assumption is it has to go to like data center buildouts. [00:07:15] Speaker A: Yeah. It's just kind of funny to. [00:07:17] Speaker C: Well, I would assume maybe like Nvidia is like, okay, it forces us to become more efficient. So if we can make our processors 5% more efficient, keep the same price, but these people are committing to buy the same number of gigawatts. We win on the deal. Well, maybe hadn't thought of it that way. It's crazy. Yeah. I like to manipulate things. [00:07:35] Speaker A: So I looked for, had a line with numbers. Let's see how you are. [00:07:39] Speaker C: Yeah. Hey, I have to build PowerPoints for work. I understand how you present the data you want. I'm sure I'm not as good as Justin, but you know, yeah, you're, you're the. [00:07:49] Speaker A: You're definitely the next runner up, I can tell you that much. All right, moving on to aws. AWS is going to become the official cloud and cloud AI partner because you got to have AI of the NBA, wnba, NBA G League Basketball, Africa League and NBA Take two Media. So there's several things I didn't know existed with this partnership. They're launching Inside the Game Powered by aws, a new basketball intelligence platform that processes billions of data points using Amazon Bedrack, Bedrock and SageMaker to deliver real time analytics and insights during live games. The platform introduces AI powered advanced statistics that analyze 29 data points per player using machine learning to generate previously unmeasurable performance metrics, with initial stats rolling out during the 2020, 25, 2026 season via the NBA app, NBA.com and Prime Video broadcasts. Playfinder technology uses AI to analyze player movements across thousands of games, enabling instant search and retrieval of similar plays for broadcasters and eventually allowing teams direct access to ML models for coaching and front office workflows. The NBA app, NBA.com and NBA League Pass will run entirely on AWS infrastructure, supporting global fan engagement with personalized in language content delivery While complementing Amazon's 11 year media rights agreement for 66 regular games on Prime Video. This partnership demonstrates AWS expanding role in sports analytics beyond traditional cloud infrastructure, showcasing how AI services like Bedrock and SageMaker can transform real time data processing for consumer facing applications at massive scale. [00:09:36] Speaker C: I mean it's interesting, which is how I'm always surprised Amazon doesn't kind of get in trouble. Here is where they're taking okay, you're going to leverage our AWS cloud, we're going to sell you on Amazon Prime Video. They're kind of, you know, bundling everything together and I'm sure, which is how they make some of these deals, which is great for them because they have that capability. But at the same point I feel like at one point we're going to get into like the 1990s, you know, you know, Microsoft antitrust because they're kind of just tackling everyone into their little home and they have to, you know, they're controlling everything at that point. [00:10:13] Speaker A: Yeah, for sure. I mean if you're Disney or I guess it's all Disney because they own ABC and ESPN and I don't know who owns tbs, but that's where I've seen basketball games before. But if you're, you know, one of those traditional providers, like how are you going to compete with all of the other bits of the platform that you don't natively run? So it is kind of, you know, one of those things. But it's, you know, we'll see how it all shakes down. I do like the AI analytics in sports like AWS is already in the NFL and F1 racing and it is kind of a neat sort of add on when they incorporate it into the broadcast. But I think that's just some of data in there. [00:10:57] Speaker C: Yeah, but I also feel like a lot of that was already there, you know, now they just branded it. You know, go back to what we call the, the topic ML is how AI makes money. You know, a lot of that stuff has been there for a long time. You've already seen Those stats of 57% of people don't, you know, don't make the kick from the 55 yard line or you know, whatever the stats are. You kind of already have that. A lot of that that was there, they just rebranded it and throw it a little bit more in your face now. [00:11:23] Speaker A: Yeah, I think I, I think it's a little bit more real time. I think there was a lot of statistics, but I think it was, you know, humans crunching the numbers like and so you had teams of people as part of broadcasts, like quickly putting stuff together. It is one of those things where some of it exists and some of it's been done by humans. And I don't know, it's a neat addition. Could be speed branding, but I think it's a couple of different things I've seen is a little bit different. [00:11:47] Speaker C: All I can think of is the story from my grandfather when they put the yellow line on the field in football. When he first saw it, he was like, how are they moving the line so quickly? And this was in the 70s, so I clearly wasn't alive. It fascinated him. Even when he went to the game, he didn't notice it wasn't there. But then on the field when they added it, maybe it was later than that, but I remember him telling me the story about this. It's kind of how I feel like some of this is, which is like cool for sure. [00:12:19] Speaker A: It's definitely the, you know, the next generation of that. [00:12:23] Speaker C: Yeah. AWS introduces self service invoice correction feature for customers that have run into this. I'm sure this has been a massive pain for them. AWS launches self service invoice correction feature allowing customers to instantaneously update purchase order numbers, business legal name addresses on their invoices without contacting Microsoft support. This must have been a pain for a lot of people for them to implement this because I've never had to run into this, thank God. But then again, I don't pay A lot of my bills directly, I let my company deal with that. And the finance teams, where you would have to open a ticket with aws, deal with the volume of and wait for them to get back to you. This guided workflow in the console lets customers update both their account settings and existing invoices, providing immediate corrections for, for prior downloads. [00:13:12] Speaker A: I want to self correct my invoice to be $0, but I noticed that that's specifically not one of the features they mentioned here. Like, so this is just changing my name at the top. I guess I could, I could point it at someone else. It's kind of the same thing. [00:13:26] Speaker C: Yeah, I mean, what I thought about first, what you said, that I thought about the same thing which was, you know, it was like, hey, we, we accidentally had this S3 bucket that was public and got DDOS in that image or we accidentally made a thing. So I thought it was going to be, hey, if it's under like $300, you know, whatever, like 1% of your bill, they'll let you automatically like, you know, fix the problem. And then I got less interested in the story. [00:13:51] Speaker A: Yeah, I mean, I'm sure it was a giant pain for, you know, those, those accounts payable teams who had to deal with like, you know, getting these corrections out. And it's already hard to, you know, issue a PO and deal with invoices. So I'm sure this is, it's the PO part, a great addition for those teams. [00:14:09] Speaker C: Yeah, it's the PO part. I can't imagine how often businesses, and maybe I'm wrong, change their legal names, I mean, addresses. Yeah, I've had to deal with that already multiple times in my career. But the POS is probably really the pain point that sparked this. [00:14:27] Speaker A: EC2 Image Builder now provides enhanced capabilities for managing image pipelines. EC2 image builder now automatically disables pipelines after consecutive failures, preventing unnecessary resource creation and reducing costs from repeatedly failed builds. A very practical solution for teams dealing with flaky build processes. Is there any other kind? The new custom log group configuration allows teams to set specific retention periods and encryption settings for pipeline logs, addressing compliance requirements and giving better control over log management costs. The update targets a common pain point where failed image builds could run indefinitely, consuming resources and generating costs without producing any usable outputs. The features are available at no additional cost in all AWS commercial regions, including China and GovCloud, making them immediately accessible through the console, CLI, API, Cloudformation or the CDK. It's everywhere. And these enhancements position Image Builder as A more mature CI CD tool for AMI creation. Competing more effectively with third party solutions by addressing operational concerns around cost control and logging flexibility. [00:15:39] Speaker C: I mean, I just like this because it automatically disables the pipeline, you know, and I feel like this is more for, hey, all those old things that you forgot about that are running that just keep triggering daily that break at one point or you hope break, so you actually don't keep spending the money on them. That's a pretty nice feature in my opinion there where just stops it from running forever. Because then you're like, okay, this thing I care about is broken. Now let me go figure out which one of these 15 failures because you. I don't have proper monitoring set up because it's an internal tool and maybe I set it up 100 correctly like I was supposed to. Don't judge me here, you know. And then the other piece is the log retention. Like I don't care about some of my, you know, hey, one off builds, but you know, maybe I care about my bricklayer builds. I need for, to keep for a year for compliance reasons. But having kind of that control and in my opinion most of the time just shrinking it down because it's a base image, you know, CloudWatch log, you know, costs a lot of money if you, you know, if you don't set that retention period and by default it's not there. So those two pieces are pretty nice, at least in my opinion. [00:16:46] Speaker A: Yeah, especially if it's just running forever and ever and ever. Right. Logging. Logging its failures. [00:16:54] Speaker C: I have the problem where like it runs indefinitely and like the EC2 insta runs indefinitely. But you know, maybe that's just a edge case I haven't, you know, ran into yet. [00:17:05] Speaker A: I mean, I, I guess I haven't seen it where the EC2 runs indefinitely, but I've definitely seen it where, you know, starts up every day, runs for oh yeah, you know, an hour and then dies at some point and then eventually will show itself down. [00:17:18] Speaker C: Yeah, I mean that's where if you're building a real platform as a service, you stop proper monitoring and alerting and do all those good things. And I try to do that all the time, but there's sometimes you're like, great, this is up. Oop. This POC went into production. Great. That has never happened to anyone on this podcast, I'm sure. [00:17:37] Speaker A: No, no, never once. [00:17:42] Speaker C: Open source MCP Server now available for Amazon Bedrock Agent Corp. AWS released Open Source MCP Server for Amazon Bedrock Agent Core that's really difficult to say. [00:17:55] Speaker A: Say that yeah, I'm struggling to say. [00:17:58] Speaker C: It once and in our notes it says model context protocol. So every time I see that, I'm like, skip that, go to mcp. We should just do the podcast about how hard it is to do what Justin does every week. [00:18:10] Speaker A: Yeah, we really should. [00:18:13] Speaker C: Anyway, I'll start this section over. AWS releases Open source MCP server for Amazon Bedrock Agent Core providing standard interface for developers to build, analyze and deploy AI agents directly in their development environments with one click installation in their IDEs. Obviously like Curo, because what else would Amazon first or Amazon Cube because those are obviously the most important. Or other tools like cloud code and cursor, which are ones that most people probably use. [00:18:41] Speaker A: Don't forget. [00:18:41] Speaker C: Jules. Oh, you think it actually integrates? [00:18:45] Speaker A: Probably. [00:18:45] Speaker C: It might. Well, it's an MCP server. That's the point of them. Yeah, it's the Swiss army knife here. MCP server enables natural language driven agent development just like every other MCP server, allowing for developers to iteratively build agents and transform agent logic into work that the agent core SDK before deploying to development accounts. Because who doesn't deploy right to production. Streamlining the path for prototype types to production. See directed to production. [00:19:15] Speaker A: Yeah. [00:19:16] Speaker C: This integration addresses the complexity of building an AI of AI agent development by providing unified protocols that work across multiple development tools, reducing friction between local development and your one person that's still stuck on Emacs or Vim or whatever, the legacy thing that they develop in and AWS deployment while maintaining security and scalability capabilities. Obviously you've never had used AI because I've had IT pull really dumb things from a security perspective. No. Might have been that NPM package the other day was the one that was vulnerable a few weeks ago. Yeah, I had something that pulled that version. I was like, oh yeah, thanks. For businesses looking to implement this, this may help reduce development time and technical barriers for your development team. [00:20:03] Speaker A: Yeah, this is one of those things where I'm a team of one right now doing like a whole bunch of like snowflake development of internal services. And so like I'm like, what, what's this for? I don't get it. This is fine. I don't understand the problem, but I can, I can imagine that this is something that's really useful more when you like spreading out against teams that you can get sort of unification on some of these things. Because if you have like a team of people all developing separate agents that are in theory somehow going to work together, which I don't think there's a lot of like, I don't think we've quite figured that out. So I think this is a step towards maybe working in that direction. We'll see. [00:20:38] Speaker C: Yeah, I don't know. I think it's another way, you know, it's another MCP server for something to talk to so you can have it build your agents for you. So you're having the agents build the AI, build the agents that other agents talk to, which will help build the AI, which will build more agents. And we definitely don't have Terminator happening anytime soon. [00:20:56] Speaker A: And then you'll have agents that you need to identify the other agents and let you know about the other agents. [00:21:01] Speaker C: At what point do we have enough agents? [00:21:04] Speaker A: We will never have enough agents. Amazon ECS now supports one click event capture and event history querying in the aws console. Amazon ECS adds one click capture that automatically creates event bridge rules and CloudWatch log groups, eliminating the manual setup for monitoring cast state changes and service events. The new Event History tab provides pre built query templates for common troubleshooting scenarios like stop tasks and container exit codes, keeping data beyond the default retention limits without requiring CloudWatch logs insights. This knowledge is a long standing pain point where ECS task events would disappear after tasks stopped, making postmortem debugging very difficult. Now an operator can query historical events directly from the ECS console with filters for time range, task ID and development ID features available in all AWS, commercial and gov cloud regions. Standard CloudWatch logs pricing for DevOps teams managing production workloads Simplifies incident response by consolidating event data at one place rather than having to jump between multiple consoles. [00:22:11] Speaker C: It's a great click Ops feature and the fact that it will actually retain old task IDs. Holy crap. That has pissed me off in the past. [00:22:19] Speaker A: Yeah, that is annoying. [00:22:22] Speaker C: That's dead. I don't know what the exit code was because you've deleted everything. [00:22:26] Speaker A: Yeah, it is the worst. If you don't have the logging fully instrumented, it's like ugh, why did you die? [00:22:36] Speaker C: You know I talk. I had a call with Microsoft today. I was like hey, we're going to leverage this service here. Can you tell me what sharp corners, you know what sharp edges this thing has? And like this was what for AW for ECs for years I write into more times than I could count. I have a lot of black and blue marks because of it. In other worlds of MCP servers, AWS knowledge based MCP server is now generally available AKA what we've Used all of aws, not replay repost to generate for you. Pretty sure that's what they've now done. AWS launched a free MCP server that provides AI agents and LLM application direct access to AWS documentation, blog post, what's new announcements and well architected best practices in a format optimized for language models. The server includes regional availability data for AWS API cloudformation resources and helps AI agents provide more accurate responses about service availability. Which is not going to be accurate because you know they change too quickly and if the service is down it's not going to tell you and reduce hallucinations. Definitely not while answering AWS related questions. No AWS account is required and available at no cost with rate limiting obviously making the accessible for developers building AI assistance or chatbots that need authoritative AWS information without manual context management. Capable MCP client or agentic frameworks that supporting the protocol allow developers to integrate Trust's AWS knowledge base into their AI application through simple endpoint configuration. This addresses a common challenge where AI provides outdated or inaccurate information because there are like 5000, 10,000 AWS what's new post a year and ensuring responses are anchored on official up to date AWS documentation best practices. Now I could see this being very useful if I used Amazon more on my day job and the JIRA MCP which I've used to automate ticket creation for people. Hey, I want to do this with this service. Go build me a, you know, build me a diagram with the well architected and kind of just start that, you know, hey, I'm building this platform as a service or something like that out and kind of get that starting point for a lot of things. [00:25:11] Speaker A: While you're reading through this, I realized that all of these MCP servers that are coming out, it's just like it's everyone's answer to the AI bots just constantly scanning the web pages and robust that tech. So it's like it's just giving them all an avenue for getting at the data. [00:25:32] Speaker C: And the key here that is it's the rate limiting here too. Like it's putting realistic controls in place for something that makes sense versus before was just they would scrape everything. Yeah, I mean we ran into this with Boltbot. Justin talked about it at one point. [00:25:50] Speaker A: Exactly. Yeah, that's like all those different rules and now if each cloud provider, you know, has mcp, you know, access to this data, then you can just use that. It's kind of cool. [00:26:01] Speaker C: It would be really nice if Azure had it, but they Also would have to consolidate all their what's new in one place because they will announce random features on random one off blog posts which is always makes me happy inside. [00:26:17] Speaker A: Yeah, I mean it makes me happy that you're managing the news feed for those. So. [00:26:25] Speaker C: That was Mead. [00:26:28] Speaker A: No, it's genuine. I'm glad that you're doing it. [00:26:32] Speaker C: I understand. I'm not. Hence that was Meadow. Well, yeah, there's a couple like newsletters, other things I. I subscribe to. I know Justin subscribes to. To try to make our lives easier. Just consolidating everything. [00:26:48] Speaker A: Yeah, well, and Azure has been traditionally the hardest to get the information for just because of how. [00:26:56] Speaker C: I did talk to somebody about it and they said it came up on discussion within one specific team a few weeks ago how it's hard to find stuff and I was like, yes. And I was like literally in the meeting with these people I was like, yeah, for my after hours thing I do with the podcast, it really sucks. So can you centralize it please? It would make my life easier. I don't know that they cared or not, but it's a different story for a different day. [00:27:22] Speaker A: We'll see when they change it. We'll take all the credit. [00:27:25] Speaker C: Yeah, 100%. It's what we do. [00:27:27] Speaker A: All right. Automatic quota management is now generally available for AWS service quotas. AWS service quotas automatically monitor quota usage and sends proactive notifications through email, SMS or Slack before customers hit their limits, preventing application interruptions from quota exhaustion. This feature integrates with aws Health and CloudTrail events, enabling customers to build automated workflows that respond to quota threshold alerts and potentially request increases. Programmatically, this is a common operational pain point where teams discover quota limits only after hitting them. What? There's any other way to cause service disruptions, failed deployments, and often during critical scaling events. The service is available at no additional cost across all commercial regions and DevOps teams. This provides like a centralized visibility into quota consumptions pattern across service, helping predict future needs and and plan capacity more effectively. [00:28:22] Speaker C: So you clearly haven't done the well architected framework review recently? [00:28:26] Speaker A: Not recently, no. [00:28:27] Speaker C: I was doing one at one point recently for somebody and this comes up in there. How do you manage your AWS service quotas? And now I really want to just respond back with I've automated it by yourself, so remove the damn question. This is not useful. [00:28:42] Speaker A: I mean it's funny because like when I was managing AWS list, we automated this too, but not in a proactive way. But by using our account management tool to just artificially inflate everything to crazy limits. You're lucky. [00:28:58] Speaker C: I think I requested a limit at one point, like they wouldn't do it. And it was like, you know, I learned never go choose like 80% of your current. Whatever your limit is, do like 80% of that. I always request that if you go over 100%, they immediately have to ask the teams. And I'm like, you're telling me going from 10, from 5 to 20 EC2 instances is like, no. Like, come on, people. Yeah, but this is definitely a nice one that does integrate you, link you up with sns. You can get the notifications. You should implement this. If you're running this at an organization, it will save you an outage, I guarantee you. [00:29:36] Speaker A: I know when they, they rolled this out on Google Cloud, I, I used it immediately just because it was one of those things. It's such a pain. And I think Google's thresholds on their service limits are very unfriendly compared to AWS in terms of instance types. And there's a lot more complexity to them because they, you know, like, not only do you have quotas per instance. [00:29:56] Speaker C: Type, isn't that the shaping feature too? [00:29:58] Speaker A: Yeah, you have all the quotas that are CPU and memory and disk, it's all separate. So you can actually increase your quota for this instance type, but then still not be able to launch because you didn't do the other thing. You're like, ah, so this is definitely something that's incredibly useful to automate. You know, when you get close to a threshold and then, you know, the, the trick is figuring out the circuit breaker, you know, for just blowing it all up. But yeah, whatever, it's just money. [00:30:28] Speaker C: I'm not the CFO. I'm good. Amazon RDS for DB2 launches support for per native database backups. And just specify that because both Ryan and I missed this. Yeah, so tell you afterwards what we thought the feature was. RDS for DB2 now supports native database level backups. The key there is level backups, allowing customers to selectively backup individual databases in a multi database instance rather than full instance database snapshots. This addresses a common enterprise need on RDS DB2 question mark. Moving specific databases between environments. Customers can now easily migrate individual databases to another RDS or backup an on prem DB2 installation using standard backup commands. Development teams can benefit from this because obviously they can move stuff around between more and it makes it more cost optimized because maybe one customer pays you for backups. And the other one doesn't and it's YOLO with their data. It's up to you. So when Ryan and I first read this, I was like, wait, they didn't have point in time and standard backups in rds? Like what? Like that was confusing. So I added the story because of that and then when I actually read the story later, it made a lot more sense. [00:31:54] Speaker A: Yeah, I think, you know, Matt and I are sharing like two, three brain cells tops this week. So we're, we're struggling through. [00:32:02] Speaker C: There was no kids sleeping this week. It's been a sleep strike in my house. Yeah, it's. [00:32:08] Speaker A: I, I don't have as nearly as good excuse, but anyway, so we're tougher game. So yeah, we both read this. [00:32:14] Speaker C: We're like, what? [00:32:15] Speaker A: How do they not have this? And then of course, you know, actually paying attention, you're like, oh, okay. That's kind of neat to be able to do it separately within the service. Does any other existed in others? I don't know. [00:32:26] Speaker C: I don't think RDS. I don't think like MySQL does or anything else. Definitely Microsoft SQL does not. [00:32:33] Speaker A: Let's be honest, I know they don't. Yeah. All right, let's move on to GCP. Gemini CLI for PostgreSQL enables natural language database management, allowing developers to implement features like fuzzy search my fuzz favorite through conventional commands instead of manual SQL configuration, extension and extension management, which I will never learn. SQL. And this is one of those features that will prevent me from having to. Which is awesome. The tool automatically identifies appropriate PostgreSQL extensions like PGTGRGM for fuzzy search. Thanks. Checks installation status, handles setup and generates optimized queries with proper indexing recommendations, reducing the typical multi step database tasks with simple English requests. Key capabilities include full lifecycle database control from instance creation to user management, automatic code generation based on table schemas, and intelligent schema exploration, positioning it as a database assistant rather than just a command line tool. It's a common developer pain. Point of context switching between coded editors, database clients and cloud consoles, potentially accelerating feature development for applications requiring advanced PostgreSQL capabilities. [00:33:48] Speaker C: I really like the potentially increasing people because they launch context switch. It's like, yeah, it's a feature. Why are we talking about context switch? Like whatever. Sorry, I don't know. [00:33:58] Speaker A: Like if, if you watch me develop database, it's. It's me on a command line going, huh? [00:34:04] Speaker C: What? Okay. [00:34:05] Speaker A: And then 15 minutes of poking around Google to get the wrong language Version or some ancient, you know, query syntax, it doesn't quite work anymore. [00:34:17] Speaker C: I think I was just banned from touching SQL on my day job. So it's better off that way. [00:34:21] Speaker A: I, you know, like I should just bite the bullet and just delete a whole bunch of data sets so they never like that's, that's the way to get at it. [00:34:29] Speaker C: You name your kid Johnny drops tables. [00:34:35] Speaker A: Yeah, I mean this is kind of neat. This is another way that you know, they're putting Gemini everywhere, integrating it across the cloud services. You know, we talked about BigQuery previously and now no Postgres SQL. So it's, it's kind of neat and I think that you know, this is going to be just the way we interact with, with data sets in the future. So that's kind of neat. [00:34:59] Speaker C: Yeah, I mean anything to make SQL better but I still worry that one, anything that makes me not have to actually do SQL, let's start with that. But then two, you know, I worry on the flip side, you know, this is where I'm gonna get in trouble. Cause I know some developers at my company. Listen, developers, writing SQL is normally not a good thing. So you. They write not always the best queries. I mean I've definitely met some that are good but we'll bypass that, you know. So I worry the same thing's kind of going to happen here where it's going to write this like ridiculously complicated inner outer union join. I'm just naming all the SQL terms I know at this point and it's going to like, you know, crash production databases. Well, the AI told you it's going to be like the AI wrote this. Maybe you should have peer reviewed it and looked at this 5000 line long SQL statement and thought maybe that's a bad life choice. [00:35:58] Speaker A: Ever heard the term out of sight, out of mind? I like that one. [00:36:01] Speaker C: That's why you have an SRE team that you're not a part of anymore. Ryan. I see how it is. [00:36:07] Speaker A: Now. I don't know why performance is bad. [00:36:09] Speaker C: It worked well in Biff. It worked on my laptop. Ship your laptop in. Yeah, exactly. Had a moment a couple weeks ago with a side project I was doing where somebody said it works here and I was like, took me every effort not to just send them the meme of like I'll ship your laptop to production. [00:36:29] Speaker A: Yeah. [00:36:30] Speaker C: This is why I questioned how I was consulted for so many years. [00:36:36] Speaker A: I think you just stored up all that anger and starting to eke out now. [00:36:39] Speaker C: Yeah, it's path of the reason to do the podcast with you guys. I got to let out that aggression. That's right, that cynicism has built up over the years. [00:36:55] Speaker B: There are a lot of cloud cost management tools out there, but only Archera provides cloud commitment insurance. It sounds fancy, but it's really simple. Archera gives you the cost savings of a one or three year AWS savings plan with a commitment as short as 30 days. If you don't use all the cloud resources you've committed to, they will literally put the money back in your bank account to cover the difference. Other cost management tools may say they offer commitment insurance, but remember to ask will you actually give me my money back? Our Chair A will click the link in the show notes to check them out on the AWS Marketplace. [00:37:34] Speaker C: Google announces a $4 billion investment in Arkansas Google is investing $4 billion between now and 2027 in Arkansas to build the first data center in the state at West Memphis. I honestly don't know where West Memphis is in Arkansas. Expanding GCP's regional presence and capability for cloud and AI workload in central U.S. dumb question. Is Walmart on Google or on Azure? [00:38:05] Speaker A: Well, we know they're not on Amazon, right? [00:38:07] Speaker C: I know. So like, is this specifically for Walmart? This investment includes a 600 megawatt solar project partnered with Energy Entergy and programs to reduce peak power uses, addressing a growing concern that AI is going to eat all the power the power grid. Google is providing free access to Google AI courses and career certifications to all Arkansas residents, starting with the University of Arkansas and Arkansas State University to build talent and local AI talent. The 25 million dollar energy impact fund for credited I don't know why I decided to read this line. County residents demonstrates Google approach to community investment alongside data center development. This positions GCP to better serve local consumers in the air with low latency options. [00:39:04] Speaker A: So a little live research. Walmart is using both Google and ashore as well as their own private data center infrastructure. [00:39:14] Speaker C: So they're on OpenStack. [00:39:15] Speaker A: Yep. Actually the article I read called that out directly. [00:39:19] Speaker C: Yeah. Oh, I made that up to be cynical, but I think it shows we've been in this industry too long when I'm making fun of things because I thought I've heard them and it's true. Yep. All right. Yeah, this one's from March 31, 2025. Why Walmart? Why OpenStack is still Walmart's private cloud of choice. [00:39:44] Speaker A: I mean, when you're that size, there's an argument to be made if you can do the economy of scale and make it work. You know, I still probably wouldn't want to do it, but it's, you know, it depends on, you know, if you can get the talent to manage OpenStack and you can keep those people happy enough where they will continue to work on OpenStack because. Oof, it's rough for my batch. [00:40:18] Speaker C: I've luckily never had to deal with that. I did talk to a couple customers that were doing OpenStack to AWS migrations and we're just going full container at that point and you know, walk them through some of the high level and everything I've heard about OpenStack is just you have to be at a scale in order to realistically manage it. [00:40:37] Speaker A: Yeah, I mean it's, it's one of those things where it's just when it goes bad, it goes bad, bad, bad, bad. [00:40:44] Speaker C: Kubernetes or OpenStack? Sorry, I got distracted with which conversation. [00:40:49] Speaker A: Yeah, no, OpenStack would, I mean both of those, that's, you could apply it to both, so that's fine. But yeah, it's, you know, and it's just got a lot of edge cases where it doesn't. [00:41:00] Speaker C: I mean it builds for a massive enterprise though, which you know, Walmart is at the scale of that so it makes sense for them. [00:41:08] Speaker A: And they have a pretty like sizable tech team, which is I, I've always found kind of surprising because I don't, you know, in my head I don't think of Walmart and technical excellence, but they're, they actually I read some really cool articles about some of the stuff that Walmart does internally. So it's. [00:41:27] Speaker C: Yeah, I always find the companies like you, I, I live in New Jersey so I fly United. But their app is significantly better, you know, than most of the other ones that I've seen. You know, and the looking back at the history, see where the planes is, you know, there's a lot of information there, you know, versus what I used to fly JetBlue also a lot. You know, there's just not a lot in there versus this. Like the app is actually really good though half the time I still want to yell at it, but it's amazing where non tech companies build really solid tech platforms and if you do it right and if you're listening to this, I assume you already know this, you really can help innovate your company. [00:42:06] Speaker A: All right, moving on to Azure. I think that's Matt's favorite. [00:42:11] Speaker C: What? Oh no. Your turn. Yes. [00:42:14] Speaker A: I know what I'm doing. Microsoft is restructuring its Commercial organization under Judson Althoff as CEO of the commercial business. Consolidating sales, marketing, operations and engineering teams to accelerate AI transformation services for the enterprise customer. Their reorganization creates a unified commercial leadership team with a shared accountability for product strategy go to market readiness and sales execution. Streamlining how Azure AI services are delivered to customers. Operations teams now report directly to the commercial leadership rather than corporates which should tighten the feedback loops between the customer needs and Azure service delivery. Structural change allows Satya, Nadella and engineering leaders to focus on data center buildout, systems architecture and AI innovation while commercial teams handle the customer facing execution. This move signals Microsoft's push to position itself as the primary partner for enterprise AI transformation, likely intensifying its competition with AWS and Google Cloud. Dun dun dun. [00:43:22] Speaker C: I mean the split kind of makes sense. You know their data centers need some love and some focus on these things and having someone focus on customer growth, you know, is probably a pretty good, pretty good, you know, idea to kind of let people kind of focus on specific target areas because I think it was just too broad before. [00:43:45] Speaker A: Yeah, I can understand that. I'm curious to see like, because it's, it's a little vague on what all is going to be in this new organization. So it's like from my non Azure use, you know, user standpoint I'm like what does that mean? Like does it mean all this? All the application development of Azure services is under one team and data center build outs are now somewhere else or you know, like that kind of thing. But we'll see. Hopefully it works out and you know, hopefully it does accelerate things or it's just the AI. It could be just the AI business. It's actually pretty. [00:44:17] Speaker C: Yeah, I think it's just the AI. Even our account team changed their names a bunch. One of them, you know, when everyone is like AI data, AI data specialist, which is really just SQL so they all have AI now in their name. And I was like cool, cool. I have these 37 questions for you. Go. They were not happy when I came back from paternity leave. I had a lot of questions. We also had a count team changeover and our meetings dropped off a few weeks before. So it was like eight weeks of stuff plus I had a month off. So I came back firing and they were unhappy with it. [00:44:57] Speaker A: Nice. [00:44:58] Speaker C: In other things I am, you know, whatever about Grok 4 is now available in Microsoft Azure AI foundry. [00:45:06] Speaker A: Segues are hard. [00:45:08] Speaker C: Segues are hard. I had a better one but I decided not to go there. Microsoft brings Xai Grok 4 model to Azure OpenAI foundry with 128k token contents window, native tool usage, integrated web capabilities positioning it as a competitor with GPT4 and Claude for Enterprise Reasoning tests. This model features Think mode for first principle reasoning that break down complex problems into step by step making them particularly suited for research, analyzation, tutoring and troubleshooting scenarios. Pricing starts at 2 million per input tokens and 10 million 10 sorry, $2 per million input tokens. That was a really big price difference. [00:45:51] Speaker A: Yeah, I was going to flip the laptop. [00:45:56] Speaker C: And $10 per million output tokens for Grok 4 with faster variants at lower costs. AI safety is enabled by default on GROQ models. Thank you for enabling security and safety by default. Also it's in your principles that say you do that, so I'm confused why they specifically called it on the article. The extended content windows also allows for entire code repository less if you have 5 million lines in your repo or hundreds of pages of documentation as single request reducing the need of manually chunking large data center to smaller ones and then having to put it all back together. [00:46:31] Speaker A: I mean I find it funny of the AA safety because whenever you hear a crazy like AI gone off the rails it usually is somehow related to grok, you know. So it is sort of funny and you know like it's. [00:46:45] Speaker C: I get. [00:46:46] Speaker A: I like competition generally and so it's good to see I guess another competitor model developer, you know. But it is like they are adding features that are kind of one model behind Anthropic and OpenAI. I don't know, I've never used it so I don't know if it has its niche or. Or use case like I've seen some. [00:47:10] Speaker C: People use it for like software development. Not so much as in like you know, Cursor or Ruko or any of those but you know, just in straight question and answer and it did provide correct answers. So you know there was that. But I've also seen it do kind of crazy things. [00:47:29] Speaker A: Microsoft is going to allow consumer copilot in corporate environments. [00:47:35] Speaker C: What? How's that security hat going right now? [00:47:39] Speaker A: Yeah, exactly. Personal Copilot subscriptions Personal family or Premium will work with work Microsoft 365 accounts effectively endorsing shadow IT practices while maintaining that enterprise data predictions remain intact. Through entre identity controls, IT administrators can disable this feature and they are rushing to do that right this second through cloud policy controls and audit. Personal Copilot interactions through the default enablement removes their initial authority over AI tool adoption within their organizations this move positions Microsoft to boost copilot adoption by any means necessary, apparently by counting the personal usage in enterprise environments. While competing AI vendors may view this as Microsoft leveraging its Office dominance to crowd all alternatives. You think government tenants GCC, DoD are excluded from this capability because they would riot and employees should note that their personal copilot prompts and responses will be captured and audited by their employers. This feature represents Microsoft's shift from preventing shadow IT to basically enabling it, potentially creating compliance challenges for organizations with strict data governance while offering a controlled alternative to completely unmanaged AI tools. I think this is nutso. [00:49:04] Speaker C: Yeah, I mean they're clearly just trying to gain adoption, but it's gonna open a pretty big. If your organization maybe has sensitives, government, anything else like that, you know, data and you might, you might not be on, you know, GCC or DoD level, you know, ASH or you know, GovCloud High or any of those you still have, you still could be leaking information which could be a massive problem. So you know, shadow IT departments are grow, going strong, never gonna end. Yeah, it's just, it's, you know, on. [00:49:41] Speaker A: One hand I'm like, well if you're, if you're an OS 365 subscriber through work, you're already partnered with Microsoft and so like sending the data to also Microsoft is probably not that big of a deal, but it is sort of an odd move where like I get it because you know, like I want to, I want to be able to automate my email, my Word docs and all that stuff. And then if my IT shop has not subscribed to Copilot because it's a bajillion dollars, you know, I can still unlock that capability for just myself and it's effectively just me accessing the same information I wouldn't for work. Like it kind of makes sense, but also just sounds crazy town. [00:50:23] Speaker C: Yeah, I mean if it's all, you know, I assume. I haven't looked at this, you know, for O365. It keeps all the data, they don't train on it. You know, the same things like Azure OpenAI, they don't train on it. It's all isolated to your tenant. You know, they have massive pages on their website about all the security they do to make sure it's not done. And that way everything else along those lines, it's at the end of the world. But if your company has a strict no AI policy or anything else like that, you are letting people have a big loophole, right? [00:50:52] Speaker A: Now, Yeah, I mean and you know it's like usually a lot of those, like no training on the data is part of the enterprise agreement and I don't know that it's necessarily going to be captured by these personal licenses, but who knows. [00:51:07] Speaker C: Fabric mirroring for Azure SQL managed instances is now generally available. Azure SQL managed instance mirroring enables near real time replication to Microsoft Fabric OneLake without an ETL process. Yay. No ETL tools supporting both data changes and schema modifications like column additions dropping unlike traditional CDC approaches. But you still have to manage pipeline a little bit. But they're saying it all. This feature provides free compute storage on fabric capacity up to 64 gigabytes terabytes free mirroring storage. We are, I am a rough person reading today. With OneLake storage charges only exceeding the free tier mirroring data becomes immediately available across all fabric services including Power BI Direct Lake data warehouse notebooks, copa allowing for cross database queries between mirror databases, warehouses and lakehouses. Microsoft positions the Zero code tool zero ETL solution competing with AWS database activity streams and GCP data streams targeting enterprise seeking simplification operational data across and reducing total ownership feels like a good way to get more, more data into their, into the lake and into Microsoft fabric here. So like they're just, they're giving these. Because there are tools, you know, that do not support. I think they already have this for Microsoft for Azure SQL. I think they might have been in beta, I don't remember or private preview, whatever it was. But it's a good way to, you know, people, a lot of customers still use Azure SQL because you can bring your own license and stuff like that. So a lot of people do use this. So it's nice, you know, pivoting point to move all your data in there. [00:52:55] Speaker A: Oh for sure. And you know, like it's because Microsoft SQL Server is so memory intensive for, for performance like being able to do large queries across, you know, data sets has always been difficult with that. So like there's, I've seen any number of ways where data replication has, you know, happened in order to sort of have your sort of app layer, you know, traditional query but then also have your sort of large reporting jobs be able to run in the background and provide that data asynchronously and you know, it's, you know, I've seen it work well. I've seen it be a complete disaster because it's, you know, it's difficult to do without making a Rube Goldberg machine that can deal with all, you know, when you think about, you know, not only do you have etl, but then you have, you know, your standard database operations like you know, just restores moving around. If you've got tables that are set to replicate and you do a restore, like what happens to those settings, like it's, it can be a nightmare. So I can see why this is very handy if you're running SQL, Microsoft SQL on Azure and then the fact that they're giving you so much for free is, is, you know, the incentive there. They know what they're doing. [00:54:10] Speaker C: Well, I was going to say, let's be honest, you're paying for at that point fabric plus you're also paying for a Microsoft SQL licensing that can support 64 terabytes of data. So they're making your money from you. Don't worry. [00:54:23] Speaker A: Oh yeah, no, I'm not worried about Microsoft going out of business anytime soon. Now generally available Azure firewall updates. The IP group limit has increased to 600 per firewall policy. This triples the previous limit of that was 200, allowing organizations to consolidate more network security rules into fewer, more manageable groups. The enhancement directly addresses enterprise scalability needs by reducing rule complexity. Instead of maintaining the thousand individual IP address across multiple policies, administrators can now organize them into logical groups like branch offices or partner networks. And the increased limit brings Azure Firewall more in line with AWS network firewall and GCPEAK's Cloud Armor, which have historically always offered more flexible role management. The primary beneficiaries are large enterprises and managed service providers who manage very complex multi tenant environments as they can now implement more granular security policies without hitting the artificial limits. While the feature itself is free. Customers should note that Azure Firewall pricing starts at $1.25 per deployment hour plus data processing charges, making efficient rule management critical for cost optimization. [00:55:37] Speaker C: Yeah, Azure Firewall isn't cheap, but it's also your IDS and ips. So if you're comparing it to Palo Alto or any of these other massive ones, the premier version is not cheap, but does give you a lot of those security things. It also can do things like the AWS one they can do and that's where that $25 per hour really comes in. And it's charged the same way everyone else is charged. So I don't think it's that big of a deal. It's just a nice quality of Life improvement here. 200 feels like a lot until you start really restricting stuff. And if you are controlling all your egress outbound and making sure only proper DNS and everything else are done. That restriction there, actually you can hit that pretty quickly at 200. [00:56:24] Speaker A: It's kind of funny. Like it does seem like a large number, but then in reality probably goes very quick. [00:56:30] Speaker C: Yeah. So in the world of DigitalOcean, we have a few good news stories this week about them. [00:56:36] Speaker A: There are other clouds. What? [00:56:39] Speaker C: There's companies we like that we call clouds? No, DigitalOcean has a couple good stories that Justin added for us, so we'll do our best to make Justin proud here. Announcing Cost Effective Storage with NFS Cold storage and Usage based Backups Digital Cloud is launching NFS, a managed file system starting at 50 gigabyte increments, supporting NFS v3 and 4 and allowing for multiple CPU and GPU droplets to mount the same share for AI ML workloads. This addresses the need to share high performance storage without the typical one terabyte minimums of their competitors. Space Cold storage enters preview at 0.7 cents per month with one month retrieval. Trying to make sure I got that right because I screwed up the one above so badly you did it faster than I did with one month free retrieval. Targeting petabyte scale data sets that need instantaneous access but are used rarely, the pricing model avoids unpredictable retrieval fees common with other providers. Infrequent access on AWS by including one month retrieval in the base price. Pretty sure AWS sort of does the same, but not really. Usage based Backup now supports 4, 6 and 12 hour backup intervals with retention from 3 to 6 months. Prices range from $0.01 to $0.04 per gigabyte per month based on frequency. This consumption based model helps businesses meet their strict RPO requirements without paying for unused capacity. All three of these services target AI ML workloads and data intensive applications with NFS for data training, cold storage for archival and frequent backups for GPU droplet protection. The combination provides a complete storage strategy for organizations with growing needs. These are services are currently available only in a couple regions but are planning to grow and if you need access please submit a support ticket. I do like how DigitalOcean still has their target little audience that they are you know really just supporting their, you know, niche market and they're they're going to stay there because a lot of companies don't need the scale, the flexibility, everything else that you know, aws, Azure, gcp, all the big providers provide and these simple things for a smaller medium sized business that maybe is just running a simple rag model on top or you know of their workload is probably all these Companies need and they're, they don't have to deal with the complexity. [00:59:18] Speaker A: Of all the other clouds for sure. And it is interesting to see like you know, AI tailored specific features. Right because it's, you know, in terms of Digital Ocean, you know, they have just what their customers need and they're pretty quick to adopt some of these, you know, with the model gardens and having the ability to sort of have a giant storage, you know, data set that you can query with multiple nodes. So it's kind of, it's interesting to see, you know, where sort of the industry is going by looking at sort of the smaller guy in the market. And another one from DigitalOcean build smarter agents with image generation, auto indexing, VPC security and new AI tools on the Gradient AI platform. Gradient now supports image generation through OpenAI's GPT Image One model marking their first non text modality and enabling developers to create images programmatically via the same API endpoint used for text completions. Auto indexing for knowledge bases automatically detects, fetches and re indexes new or updated documents from connected sources into OpenSearch databases, reducing the manual maintenance for keeping AI agents knowledge current. The new VPC integration allows AI agents and indexing jobs to run on private networks within DigitalOcean's managed infrastructure, addressing the enterprise security requirements and removing the exposure of those services to the public Internet. Two new developer tools are coming. The Agent Development Kit ADK providing a code first framework for building and deploying AI agent workloads. While GENIE offers VS code integration for designing multi agent systems using natural language. This will allow DigitalOcean to compete more directly with major cloud providers in the AI space by offering multimodal capabilities, enterprise security and developer friendly tooling for building AI applications. [01:01:20] Speaker C: Again I mean I feel like we're going to say the same thing about a lot of these but like they're really flirting about their audience like you said and they're going to just build specific to what their specific customer needs, you know and they, they've determined that their customers need these, you know, extra, you know, image generation, AI features, you know, so that it's just, it's a good next step for them. They, they're not always the fastest but they get there and they get it pretty well done. [01:01:48] Speaker A: Yeah and the big one to me is the VPC integration because it's, you know, you want to build data sets on, you know, data that's probably not public, you know, and especially agents for internal business stuff is really powerful. So it's having that Capability I think will unlock a lot of things for. [01:02:07] Speaker C: For customers DigitalOcean is now announcing per second billing new droplet plans BYOIP and NAT Gateway Preview to reduce scaling costs, but clearly they've never used a NAT Gateway before. DigitalOcean is switching from hourly to per second charges for droplets starting January 1st with 60 second minimum charges, which I feel like is kind of the standard now. Though I do remember when the clouds were like one hour minimums and I was like that's why they're. You know I wrote some customs ridiculous stuff to like for auto scaling groups back in the day. Scale the one that is close to the end of the hour, things like that. This could dramatically reduce cost for really short workloads AKA CI CD pipelines where you used to have to pay for full hours when you were only using minutes. The new droplet sizes help bridge the gaps between shared and dedicated CPU plans, allowing for in place upgrades without IP or data migrations. And that's a pretty nice feature too. Change names and size, Call it a day. New plans include 5x SSD variations for CPU optimized droplets and 6.5x CPU variants for general purpose. Addressing the previous large cost jump between tiers. BYOIP is generally available with a seven day setup time compared to one to four weeks for hyperscaler. And if you've ever had to move your own IP addresses, boy was this that painful. I think they made it easier now, but it was painful. This allowed for businesses to maintain their IP reputation and avoid breaking clients allow list because everybody knows the best way to do security is lock it down to just single IP addresses. VPC NAT Gateway enters public preview at $40 per gigabyte but does include 100 gigabytes of bandwidth supporting up to 500,000 simultaneous connections and that is actually pretty impressive. Managed service services provide centralized egress with static IPs for private resources without complexity of self managed NAT instances. $40 per month is pretty expensive though, but the 100 gigabytes may be balanced out. These upgrades target cost optimizations, migration friction points and particularly benefiting ephemeral workloads, auto scaling applications and business needs to maintain IP addresses during migrations. [01:04:36] Speaker A: Yeah, I mean this goes along with theme of the last couple stories, right? [01:04:40] Speaker C: This is should emergencies once together. Here's all the digital Ocean updates. [01:04:45] Speaker A: Yeah, DigitalOcean gets features that have been running for a while. I mean I remember when Persic and billing was rolled out in AWS like it changed my life because it was. It was one of those things like I spent so much time and energy managing the workloads to keep within, you know, to gain the efficiency of getting the full hour and it just wasn't worth it. So. And the amount of just mental load that went away when they went to per second billing, it was crazy. [01:05:16] Speaker C: Yeah. [01:05:16] Speaker A: It was like oh I just don't care anymore. And now spin it up, let it spin down, let it auto scale, let. [01:05:22] Speaker C: It not did you did you know if Microsoft ever went to per second or permanent billing? Because I remember with the licensing for Windows for a long time it still was one hour if you spun up a Windows server. I don't know if they ever switched that though. [01:05:37] Speaker A: I think it depended on whether you brought your own license or not. [01:05:40] Speaker C: That makes sense. [01:05:41] Speaker A: Also weird. Yeah, so it can get strange, but I don't remember because I don't use a lot. I try not to use Windows for my workloads whenever possible. [01:05:52] Speaker C: I don't blame you on that one. I'm trying to see Per second building is available for Windows but varies by instance type, but pricing varies. So according to Gemini they do have per second billing for Windows EC2 instances. But I still don't know that I believe that because I could very easily see it hallucinating on that one and. [01:06:15] Speaker A: Our last story Introducing Snowflake Managed MCP Servers for Secure Governed Data Agents Snowflake is introducing services that enable secure data agents to access enterprise data while maintaining governance and compliance controls. This addresses the challenge of giving AI agents access to sensitive data without compromising security. The MCP protocol, originally developed by Anthropic, allows AI assistants to interact with external data sources through a standardized interface. Snowflake's implementation adds enterprise grade security layers including authentication, authorization, and audit logging. Data agents can now query Snowflake databases, run SQL commands, and retrieve results without requiring direct database credentials or exposing sensitive connection strings. All the interactions are governed by Snowflake's existing role based access controls and data governance policies. This integration enables organizations to build AI applications that can answer questions about their business data while ensuring compliance within data residency. Privacy regulations and internal security policies are maintained. This managed service handles infrastructure complexity and scaling automatically. Developers can connect popular AI frameworks and tools to Snowflake data through the MCP interface, reducing the complexity and allowing for secure data pipelines for AI applications. Yeah, yeah. [01:07:41] Speaker C: I mean Snowflake has MCP now everyone gets mcp. [01:07:45] Speaker A: Yeah, I was trying to think of like I don't have any current workloads that use Snowflake. [01:07:50] Speaker C: Like. [01:07:52] Speaker A: But I, I, you know, I have Reviewed it and used it a little bit in the past. So I can kind of see that this is kind of a, you know, an enablement if you have a whole bunch of data that you want, you know, AI to be able to answer from. Allows an easy way to do that without having it do crazy things like build a whole bunch of complex SQL queries. [01:08:12] Speaker C: Yeah, I mean, at one point it's going to almost be, you know, we're announcing all these things and it gets. The per second window I did confirm, by the way, is per second, but the same point it is, you know, it's at the point where, you know, at what point is it just assumed that everyone has an mcp? You know, I feel like it'll be. [01:08:31] Speaker A: The new REST API, right? [01:08:32] Speaker C: It's the new REST API. The per second. It's the back, you know, it's the standard things that you just expect. And it's kind of where I feel like we're, we're getting close to, you know, we live on, you know, the edge. I don't know, I can say the bleeding edge, but the edge of. I feel like a lot of tech and that's why we all like to do it. And I assume people listen to us, so you hear about this stuff. But we're at the point now where I feel like this is just approaching table stakes. Weird enough to say. [01:09:00] Speaker A: Yeah, I think pretty shortly it'll be, you know, like, we used to make fun of tagging not being, you know, announced with the initial product rollout. Like, we'll make the same jokes like, oh, it doesn't have support MCP right off the bat, like, what's wrong with you? Yeah, because, yeah, I think it is going to be one of those things where it's like, yeah, you'll have the REST API for the service and the protocol for interacting with the service, like immediately. And it's, you know, AI is going to take over everything. So it's just how things are going to work, I imagine. I wonder if Rest API eventually goes away. [01:09:35] Speaker C: You just. Well, there's also. Oh, my brain just died. What's the other API that I swear I've tried to figure out? I mean, I've only tried once. [01:09:45] Speaker A: Soap. [01:09:47] Speaker C: No, hold on. [01:09:49] Speaker A: Oh, graph. [01:09:51] Speaker C: Yeah. No, yes. Yes. I don't know. [01:09:57] Speaker A: If it's not that, then you've come up with some crazy obscure one. [01:10:00] Speaker C: I think it's Graph API. Maybe it's Graph API. GraphQL. Yeah, that one hurts my brain sometimes trying to figure out with all the permutations and other things. Like what, what, what are you asking me to do? [01:10:14] Speaker A: So, yeah, no, I, I, I, I like JSON objects, I like indexing and, and variable types. And then yeah, the minute you go to graph and it's, you're, you're giving me sort of a semi relational view of data in the API return. Like I, I see the, the power behind it. It's just I have a really hard time designing towards it. Yeah, I think some of that is that I don't really build anything that would benefit from it. So yeah, I'm usually writing code that's talking, making two other things talk to each other. Not really, you know, kind of forming relations. But yeah, it is hard. Well, at least for our brains, considering how little sleep we run on at all times. And we can barely record a podcast. [01:11:04] Speaker C: Getting all four of us in a room together. Last week I was impressed. It's been like six months. I mean, that was for other reasons, but we made it last week and we made it through this week. Ryan on sleep deprived CloudPod talk host. [01:11:18] Speaker A: I'm proud of us. I mean, I do not want to listen back to this and I apologize to all our listeners, but if you. [01:11:24] Speaker C: Made to this point, I question your life choices. [01:11:30] Speaker A: Yeah, yeah, drop a note in the Slack. I gotta see it. [01:11:33] Speaker C: Yeah, we'll have Justin mail you. If you made it to this point, we'll have Justin mail you something. Yeah, he's not here to defend himself. [01:11:41] Speaker A: That's right. We can assign him for all the work. That's the whole. Go on vacation. I dare you. [01:11:48] Speaker C: All right, we'll talk to you next week. [01:11:51] Speaker A: All right, Another day in the week in the Cloud. Bye, everybody. [01:11:54] Speaker C: Bye, everyone. [01:11:58] Speaker B: And that's all for this week in Cloud. We'd like to thank our sponsor, Archera. Be sure to click the link in our show notes to learn more about their services. While you're at it, head over to our [email protected] where you can subscribe to our newsletter, join our Slack community, send us your feedback and ask any questions you might have. Thanks for listening and we'll catch you on the next episode.