Episode Transcript
[00:00:00] Speaker A: It's really hard to optimize something if you don't know who owns what and what it does. What percentage of your applications do you have engineering headcount to redesign and rearchitect? Right. And we just continued to build and.
[00:00:09] Speaker B: Build and build more features because they.
[00:00:11] Speaker A: Were, they had such high ROI and helped organizations really get the most out of cloud spend. You can use AWS efficiently and effectively and it's going to scale with you and it's going to provide you the opportunity to change your cost structure over time to serve your business. And what can we do to accelerate finops with AI?
[00:00:30] Speaker C: Hi everyone and welcome to a new episode of our podcast. Today we have a very special guest that you will probably know from your updates, especially if you use aws. But first of all, I want to present my co host, Damian Damien. How are you?
[00:00:47] Speaker D: I'm doing great, thank you for asking.
[00:00:49] Speaker C: How are you? Doing well, doing well and today I'm super happy to have Rick Oaks here. How are you Rick?
[00:00:56] Speaker A: Doing great.
[00:00:57] Speaker B: Thanks for having me on the podcast.
[00:00:59] Speaker C: Yeah, so this guy is making me work late when I make my newsletter issues and today we are going to make him some questions that are probably going to have fun. So Damien, go ahead and start.
[00:01:13] Speaker D: First of all, I think that we would like to know how you got into Phenox. How do you feel about it first time and how you get into base?
[00:01:23] Speaker A: Yeah, it, it was sort of by accident. So I've been doing jobs I think for a little over 10 years now.
I was, I was at Microsoft, I spent 13 years at Microsoft and I had kind of started as a build engineer, kind of going up through the ranks to do resource management, IT asset management and then build. I was doing a lot of build engineering and I had sort of matured in my career up to the point of owning data centers and I was.
[00:01:49] Speaker B: Doing a data center migration project kind.
[00:01:51] Speaker A: Of as a project manager and halfway.
[00:01:54] Speaker B: Through our migration project to move to.
[00:01:55] Speaker A: A new data center, Azure became really popular and we're like okay, IT leadership.
[00:02:01] Speaker B: Wanted us to stop moving everything to.
[00:02:04] Speaker A: New data centers and move to the cloud instead. So we had this big lift and shift project that they, they wanted me to project manage and I started migrating a significant portion of non production workloads to cloud.
And as I'm migrating these things, I'm.
[00:02:21] Speaker B: I'm looking at these VMs, these instances.
[00:02:23] Speaker A: And I'm like wow.
[00:02:24] Speaker B: I, I remember building these five, ten.
[00:02:25] Speaker A: Years ago early in my career when I was a junior build engineer and I'm like, nobody's using these. This is a waste of money.
[00:02:32] Speaker B: This is so inefficient.
[00:02:34] Speaker A: And so I went to leadership and.
[00:02:36] Speaker B: I'm like, hey, we shouldn't be migrating these. This is not an efficient use of time.
[00:02:39] Speaker A: And they're like, hey, we have lease renewals. We have to get out of these.
[00:02:42] Speaker B: Buildings by a certain time.
[00:02:43] Speaker A: Just go, just do it anyway. And I'm like, all right, but I want you to promise me after this.
[00:02:47] Speaker B: Project is over, I can do an audit and build a dashboard to show all of the waste.
[00:02:51] Speaker A: And they said, okay, fine. So we migrated out of the data center. We got out with like 12 hours to spare before the lease renewal was due. So it was pretty hectic. And then. And then, true to their word, they.
[00:03:02] Speaker B: Supported me in building a dashboard and set up reports to show waste and. Right. Sizing opportunities for VMs.
[00:03:08] Speaker A: And this was like 2014 time frame. So I built this dashboard and went around to the different business units and worked with them to understand the data.
[00:03:17] Speaker B: And to go take and capture that.
[00:03:19] Speaker A: That waste, that money. And it was extremely successful. We saved over, you know, I don't know how many figures, nine figures in. In a month or, sorry, in a year.
And it was so successful that we.
[00:03:36] Speaker B: Started adding more features to the dashboard.
[00:03:38] Speaker A: And, and more visibility and more recomm and ways to save money. We implemented a project called Screen Test.
[00:03:46] Speaker B: Where it would automatically turn off servers if they hadn't been used for 30.
[00:03:49] Speaker A: Days and things like that. And it got me so excited. And we just continued to build and.
[00:03:54] Speaker B: Build and build more features because they were. They had such high ROI and helped.
[00:03:58] Speaker A: Organizations really get the most out of cloud spend.
[00:04:01] Speaker B: And then I sort of like, look.
[00:04:02] Speaker A: Back on that time and I was like, wow.
[00:04:03] Speaker B: I was kind of doing product management.
[00:04:05] Speaker A: Even back then, but also leading a FinOps practice before the FinOps board even existed.
And so that matured and I went.
[00:04:14] Speaker B: And I joined turbomnic and I spent four years there redesigning and working on their, right, sizing engine and the automation capabilities there and then.
[00:04:21] Speaker A: And I've been at AWS leading the optimization product team for the last three years or so. And so the journey has sort of matured and extended. But what's interesting is it's the same problems. We're still talking about right, sizing, still talking about recommendation, still talking about waste cleanup and 10 years in the same conversations. But I mean, they have changed maturity and execution, but the problems haven't really changed. So it's fascinating.
[00:04:52] Speaker D: From 2014, wow. Five years before the Fino foundation started and talking about the Finos foundations today, I found out that you have a role there. Can you tell us a little bit about it and, and what do you think?
[00:05:08] Speaker A: Absolutely.
[00:05:08] Speaker B: I sit on the technical advisory council.
[00:05:10] Speaker A: Of the foundation and in the council.
[00:05:13] Speaker B: We all meet on a regular basis.
[00:05:15] Speaker A: To talk about the framework, to look at feedback, look at where the industry is going.
The council has representatives from all the major cloud providers, it has practitioners, it has representatives from a few of the big third party products as well, the.
[00:05:34] Speaker B: FinOps products as well.
[00:05:35] Speaker A: So we want to make sure that it has like a really well rounded.
[00:05:39] Speaker B: Set of opinions in there so we.
[00:05:40] Speaker A: Can all really provide tons of checks and balances and have a lot of opinions that we can figure out and make sure we're serving the practitioner communities.
[00:05:50] Speaker B: It's a fun role.
[00:05:51] Speaker A: I really enjoy it. I enjoy the foundation a lot. I think the folks in the foundation.
[00:05:56] Speaker B: Are fantastic and just a pleasure to work with.
[00:06:01] Speaker C: No, that's, that's right. And, and since you have this role and based on your large experience of well, doing this Phenox practice, even though it wasn't called that way before, where, when you were doing it, what do you think is going to be the problems of the near next five years? For example, in, in Phenom, because you said the, the problems are repeated all over again and the, the right sizing opportunities and, and all this stuff, it's been there forever. But do you think it's going to be new problems and if you do so, which, which ones are going to appear in the next, I don't know, three, five years?
[00:06:39] Speaker A: I think, I think optimization will evolve. I think the core parts of it.
[00:06:44] Speaker B: Will have a lot of similarities in the next five years. I think we'll still see recommendations, I.
[00:06:49] Speaker A: Think we'll need to continue to see recommendations evolve as far as context and accuracy goes.
Right now, recommendations are technically accurate.
[00:06:58] Speaker B: They, from a mathematical and scientific perspective.
[00:07:01] Speaker A: We'Ve gotten to the point where they're, they're not wrong. They won't hurt things right From CPU memory, disk network, IO throughput inventory, CPU performance metrics.
[00:07:12] Speaker B: These are all in our right sizing models now.
[00:07:14] Speaker A: And we've gotten to a point where an engineer looking at a recommendation can't really point to a reason why they're.
[00:07:20] Speaker B: Wrong unless there's context that is not factored in.
[00:07:24] Speaker A: Say I'm running SAP and it has.
[00:07:26] Speaker B: A certain instance type that it requires.
[00:07:29] Speaker A: Or if you have other business requirements.
[00:07:31] Speaker B: Or application context requirements that need to be factored in.
[00:07:34] Speaker A: So I Think that's a big part of it.
[00:07:36] Speaker B: Automation is certainly something that is going.
[00:07:39] Speaker A: To continue to evolve. We've had automation in the industry for.
[00:07:42] Speaker B: A long time now. Again, back to Turbo. Turbo has been doing automation for years.
[00:07:45] Speaker A: And years and years. I think we're going to start seeing automation become a little bit more adopted.
I think as more companies get past the visibility maturity curve and really start.
[00:07:59] Speaker B: To understand their unit costs or cost.
[00:08:01] Speaker A: Per business, you know, whether that's business unit or app team or org structure, what have you, that visibility is required. It's really hard to optimize something if you don't know who owns what and, and what it does, right? And as more companies find that maturity.
[00:08:15] Speaker B: On visibility, we're going to see optimization.
[00:08:17] Speaker A: Really gain some more lights. And I think a lot of companies to circumvent this need for technical understanding of these resources and architecture side, they really focus on commitment optimization in just great adoption.
[00:08:33] Speaker B: Maturity and commitment optimization for savings plans.
[00:08:36] Speaker A: And RIs and all of these sorts of things.
But really that's only half of the puzzle, right? Resource optimization, utilization optimization, architecture optimization, the amount of savings and cost efficiency we.
[00:08:49] Speaker B: Can get out of that side of.
[00:08:51] Speaker A: The optimization game is, is comparable there. I, I consider them kind of two halves of the whole and we have.
[00:08:59] Speaker B: To do both together.
[00:09:00] Speaker A: And so you know that five year.
[00:09:02] Speaker B: Vision is a more complete and mature.
[00:09:05] Speaker A: Set of resource utilization optimization capabilities and seeing customers really take that more seriously and have programs that are much more.
[00:09:15] Speaker B: Focused on end to end resource configuration and optimization.
[00:09:20] Speaker A: And it's not just like deleting unused resources. It's about hey, there's 900 instant sizes.
[00:09:27] Speaker B: In AWS, which one are you going to use? And it's not just go to the next generation.
[00:09:32] Speaker A: It's a deep strong understanding of the.
[00:09:35] Speaker B: Performance differences of each family and the right fit.
[00:09:39] Speaker A: So there's a lot of maturity to go get.
And again, like we said, like same conversation, which instance type do you use?
[00:09:47] Speaker B: It's the same question we've been asking and attempting to answer for a decade now.
[00:09:52] Speaker A: And I'm excited to move past it. This is one of the things we tried to do. You know, we launched a bunch of stuff in the last few weeks. We launched in Compute Optimizer a bunch.
[00:10:02] Speaker B: Of new IDLE recommendations.
[00:10:04] Speaker A: This one was fun because the 2024.
[00:10:07] Speaker B: Survey data from the FinOps foundation, the.
[00:10:09] Speaker A: Number one issue was or focus area for practitioners was waste cleanup. And then number two is commitment management. So what did we launch this year?
[00:10:18] Speaker B: Number one, we launched a new consolidated.
[00:10:21] Speaker A: Idle cleanup capability and compute optimizer.
[00:10:23] Speaker B: And number two, we launched a brand.
[00:10:25] Speaker A: New savings plan analyzer commitment tool to do all of the analysis for you on your commitments.
[00:10:31] Speaker B: So boom. The top two most important and, and.
[00:10:35] Speaker A: You know, largest amount of time spend for FinOps teams.
[00:10:40] Speaker B: We just launch products so hopefully the next survey. Those aren't the top two things that.
[00:10:44] Speaker A: They focus on because we've automated them.
[00:10:46] Speaker B: And we built tooling around them. Right.
[00:10:49] Speaker D: That's excellent. I mean I, I love the, the, the, the things that you are doing and also to understand a little bit about the problems and that you are tackling the questions and you're totally right.
Always the same question.
I still think that there are companies that still don't understand that they need this oil rather the Phenosol they are trying to, they don't see the need to hire one or hiring company like they are fighting it or they think if we can get a tool or something like that it's enough.
This is one point that I hope that it will change.
And what you say also about the commitments of continent first to do right. Sizing right. Or, or do real potential before you commit because then you will be either over committing or committing wrongly.
[00:12:01] Speaker A: Like I, I, I tend to tell.
[00:12:03] Speaker B: Customers to do both at the same time.
[00:12:05] Speaker A: Just mix and match, right. Buy a little bit of your commitments, take a little off the top while you're doing your architecture work.
[00:12:13] Speaker B: Right.
[00:12:14] Speaker A: If you wait for a year to.
[00:12:15] Speaker B: Do your commitments, you're going to pay on demand for a year.
[00:12:17] Speaker A: And that's very expensive ways. There's ways because like in a perfect world you would right.
[00:12:23] Speaker B: Size everything, re architect everything and then commit.
[00:12:26] Speaker A: But what percentage of your applications do.
[00:12:28] Speaker B: You have engineering headcount to redesign and rearchitect?
[00:12:31] Speaker A: Right.
[00:12:31] Speaker B: Like right.
[00:12:32] Speaker A: It's not even more than 10% usually for large companies. So you gotta do both. You gotta mix and match and go carefully.
[00:12:41] Speaker C: Right, right, right.
[00:12:43] Speaker D: You know, totally right there. You start slowly with maybe saving times and focusing on big projects. And we always talk about rowing, right? It's all about roi. That's why we need to put in now because like you say a reality taking might take a lot of energy or resources.
But yes, that's a good point.
But thinking about what would you tell these companies that are still thinking that pinot is required or not?
[00:13:15] Speaker B: I'm actually okay.
[00:13:16] Speaker A: Like I run into this too.
I'll talk to companies and they're like you know, we don't have a cost problem. And I'm like okay, I'm glad to hear I, I appreciate that perspective and I'll probably talk to you next year.
Okay, there's the shift left conversation is.
[00:13:35] Speaker B: Interesting because how do you shift left.
[00:13:37] Speaker A: If you don't know you need to.
How do you, how do you think.
[00:13:40] Speaker B: About cost as a resource or spend money dollars as a resource from an engineering perspective?
[00:13:46] Speaker A: You don't know you have cost pain yet. And the whole shift left thing and getting in front of FinOps is, is important but until you know it's a business problem for your organization, it is.
[00:13:58] Speaker B: Really tough to prioritize without signals, without an understanding of okay we're over budget.
[00:14:03] Speaker A: Or hey, we don't know what our ROI is. And you know I've, I've had a lot of conversations over, over the roles and years and companies I've been at where you know we're like oh yeah, we don't really need to think about optimization and, and I've had other companies hey how to optimize before everything and it's like well you know that, that analysis is difficult and I've also seen the opposite where companies will freeze and.
[00:14:29] Speaker B: They'Ll do so much analysis for years how to redesign everything to cloud native.
[00:14:35] Speaker A: That they don't move anything.
And so both are bad. I, well I can't say bad. Both, both have you know, pros and cons that. Well, let's say it that way. Let's say that way because you can have paralysis by analysis. You can just sit there and kind of, you know, strain against the problem.
[00:14:55] Speaker B: For a long time and not make.
[00:14:57] Speaker A: Any progress failing fast. I'm, I'm a huge fan of and learning as you go and because that's what creates expertise, that's what creates heroes in your organization. People that step forward and, and do.
[00:15:12] Speaker B: The work to learn the lessons and feel the pain.
[00:15:16] Speaker A: Right?
[00:15:17] Speaker C: Yeah, definitely. You cannot be making the mistake of the analysis paralysis. You have to iterate fast and fail fast and learn because a lot of, even though we are engineers and we try to do things exactly and precise a lot of things are a B testing and see if it works or not and going to the hands on work and yeah that's, that's a very interesting thing. And, and I also can relate on the, on the commitment versus Life thing. I had an experience with, with a company that made probably one of the best commitment exercises I have seen yet. They had a major problem with S3 buckets that was super dummy to do like it it just needed to put a life cycle rule. I was that's it that they wouldn't need almost anything to do. Even though on the commitment side they were like a perfect combination of reserve instances and savings plans. And I was like, I never, I could never do it better. I was like, yeah, I take it was amazing.
[00:16:19] Speaker A: Yeah.
I mean that's why we built the cost optimization is to pull right sizing.
[00:16:24] Speaker B: And commitments together into a single screen to show people that you need both.
[00:16:29] Speaker A: The amount of savings you get from Right.
[00:16:31] Speaker B: Sizing can be pretty similar to the.
[00:16:32] Speaker A: Amount of savings from commitments. So do both. Right.
[00:16:37] Speaker C: So now that you're talking about this, and I'm curious to see, and I'm not sure how much you can tell, but this is probably fun.
So when people in the AWS teams, you know, in the AWS services release a new service with all the new costs and all the new structure and all new services, even though in Rainbed it seems to be noticed that there wasn't any announcement in terms of new services. So they are maturing new things, which I think it's also great. But when a new service comes, how do you approach it from your team.
[00:17:12] Speaker A: That'S happy to share a little bit about how my team thinks about prioritization, how we think about the experience customers have when it comes to configuring, you know, services most optimally. I would say that like from a.
[00:17:29] Speaker B: Very high level perspective, my team is.
[00:17:31] Speaker A: Not about savings, it's actually about performance.
And this gets lost a little bit.
[00:17:36] Speaker B: Sometimes because we talk about cost so much.
[00:17:39] Speaker A: But we actually want all of our customers environments to first be high performance.
And the whole point of running your.
[00:17:49] Speaker B: Application in cloud is that it can.
[00:17:51] Speaker A: Scale, it can meet customer demand spikes that maybe were unplanned or unexpected.
And so all of our tools, Compute Optimizer, Cost Optimization Hub, like all of all of the recommendations we build, they're actually all of the analytics we do in the science and the models and the ML work is all about performance. So we study exhaustively the CPU performance differences between different models.
We nerd out on that.
[00:18:20] Speaker B: We've got some incredible scientists in our.
[00:18:22] Speaker A: Organization that spend a ton of time on this. And we're looking at benchmarks.
[00:18:26] Speaker B: We do our own benchmarks.
[00:18:27] Speaker A: We compare them against real world usage numbers. We focus so much on performance and it just so happens you save money when you focus on performance because you.
[00:18:37] Speaker B: Use faster processors and you can lower.
[00:18:39] Speaker A: Processor counts and still get better latency and better performance out of your apps with less cores. And so you see a lot of savings out of our products.
[00:18:50] Speaker B: But the recommendations are actually how to Provision correctly for max performance.
[00:18:55] Speaker A: So Compute Optimizer. You know, sometimes I get on a call with customers and they call it Cost Optimizer. And I'm like, no, no, no, it's Compute Optimizer.
[00:19:06] Speaker B: It optimizes the compute performance of your.
[00:19:08] Speaker A: Resources and it just so happens to.
[00:19:10] Speaker B: Save you money by not blindly over provisioning.
So when we talk to customers, we're.
[00:19:16] Speaker A: Thinking first about how do we serve their needs by, you know, providing services and products and capabilities so they can run a great app. That's the first goal. The second goal is, hey, you can use AWS efficiently and effectively and it's going to scale with you and it's going to provide you the opportunity to change your cost structure over time to serve your business.
And when a new service is launched or a new feature is launched, we talk directly to customers. How are you configuring it?
[00:19:51] Speaker B: What's important to you when you're thinking.
[00:19:52] Speaker A: About how to build with this service?
[00:19:56] Speaker B: Is there an opportunity to mistakenly over provision?
[00:19:59] Speaker A: Is there an opportunity to set a.
[00:20:01] Speaker B: Policy boundary that's incorrect or leads to unexpected cost increase?
[00:20:07] Speaker A: And so we look at how easy is it to fix that? Is it something that, you know, engineers need help with? Is it something that, if we were.
[00:20:16] Speaker B: Building a recommendation, it would be easy.
[00:20:18] Speaker A: To take that recommendation? Because there's a big difference between a recommendation that requires two or three clicks to execute versus one that takes the RE architecture of an entire product. So we think about that exhaustively.
[00:20:30] Speaker B: We think about, hey, what percentage of.
[00:20:32] Speaker A: The customer's cloud bill is impacted by this resource?
And then we also think about, hey, when we work with this service team, like, is this service team really experiencing feedback from customers where they're like, hey.
[00:20:50] Speaker B: This is a great service, it scales.
[00:20:51] Speaker A: Really well, but we have low visibility on cost. And then we really like working with the service teams. Especially, like when those service teams sometimes come to us and they're like, we want to build a recommendation, we're like, come on in.
[00:21:03] Speaker B: Like, let's, let's work together.
[00:21:06] Speaker A: And so we get the privilege of sort of working across all of aws. And I mean, we built the new idle recommendations of the EBS team, the ECS Fargate team, right? Like, so we work with the RDS team, the Aurora team, we work with all of these different service teams.
[00:21:23] Speaker B: And my team will bring the finops.
[00:21:26] Speaker A: Perspective and they'll bring the service operations perspective and the advanced understanding of how the service works.
[00:21:32] Speaker B: So we'll work with their principal engineers.
[00:21:34] Speaker A: We'Ll say, hey, how are your Customers configuring these resources.
[00:21:38] Speaker B: How do we provide a Recommendation that.
[00:21:40] Speaker A: Is 100% safe, that is not going to break anything? And so they were a bit of.
[00:21:45] Speaker B: A check and imbalance with each other. Right.
[00:21:47] Speaker A: Where they can check us from a.
[00:21:50] Speaker B: Quality and accuracy and safety standpoint and.
[00:21:52] Speaker A: We'Ll bring that finops perspective.
[00:21:55] Speaker C: Yeah, that's super interesting because I was, when I. When you were talking about. And I was thinking about the new services that come year after year and because we had this conversation about with practitioners that. Yeah, you have, I don't know, I think it was 700 types of easy instances that you can actually configure differently and all this stuff. And I was super interested on how you approach it. And I'm also prone to see like your customer obsession in general, especially in aws, because I was in the optimization day here in Madrid and it was like obsessed with feedback and how customers do it. Because I always think that it's easier to build from the customer feedback as well. It's way easier to do the products to the customer when you have customer feedback. Otherwise it's guessing.
[00:22:43] Speaker A: Yeah.
It takes a commitment though, because, you know, it's. It's not hard to look at your backlog, look at data already in your hands and prioritize a backlog.
[00:22:57] Speaker B: Right.
[00:22:57] Speaker A: There's a lot of prioritization frameworks out there from a product manager perspective, but meeting with customers and seeing the frustration on their face or seeing the happiness on their face and converting that to data.
[00:23:11] Speaker B: Right. And we have a product feature request.
[00:23:13] Speaker A: System where our account managers, our technical account managers and our solutions architects can feed that feedback into us kind of at scale across our whole customer base, not just gathering that data and customer names and hey, who's struggling with what, but actually like sitting down and empathetically listening to, hey, you're a FinOps organization.
[00:23:37] Speaker B: You are trying very hard to influence.
[00:23:39] Speaker A: Your engineering teams to consider cost as a resource. Where is that working and where is it not working?
What data sets do you need to make this argument more convincing?
[00:23:51] Speaker B: What data sets do you need to get engineers to take. Right.
[00:23:54] Speaker A: Sizing recommendations more seriously or to understand.
[00:23:57] Speaker B: That they're safe and they're not?
[00:23:59] Speaker A: Risk, you know, you're not going to.
[00:24:00] Speaker B: Incur risk by taking these recommendations. What do you need to empower your success?
And if we can build a product.
[00:24:07] Speaker A: That makes that Finops practitioner two times more successful in their job, two times more impact on counting savings, getting to a unit economic number or something like that, that's really the success of that.
[00:24:21] Speaker B: FinOps team is the success of my products.
[00:24:24] Speaker A: Right? It's one and the same. And it's interesting because we talk about features and data sets and recommendations, but at the end of the day, finops is much more of a cultural movement and it's a philosophical conversation on what.
[00:24:40] Speaker B: Does it take to instill bravery in engineers to make changes to the way.
[00:24:45] Speaker A: They do things and how do you pivot culture. And ideally you have the right products and data sets to have a culture conversation and pivot on culture.
But the culture part is harder. The culture part is harder than the software part, than actually building the features.
[00:25:03] Speaker B: But they have to work together.
[00:25:04] Speaker A: You have to have the right products, the right data, the right experience to make this culture conversation easier, to inspire people to build a product people enjoy using.
[00:25:14] Speaker B: And they share with their coworkers, they share with their peers in the industry.
[00:25:17] Speaker A: And they're like, hey, I was really successful with this and this helped me.
[00:25:20] Speaker B: Really convince people that this approach to.
[00:25:22] Speaker A: Savings plans is the highest ROI or this approach to right sizing is, is.
[00:25:27] Speaker B: The most effective and the fastest to.
[00:25:29] Speaker A: Results or whatever that is. The products must inspire that confidence to.
[00:25:34] Speaker B: Support the cultural Change and the FinOps.
[00:25:37] Speaker A: Teams to really help influence engineering at scale.
[00:25:42] Speaker C: So super interesting, interesting reflection on, on that and I'm curious following on the, even though it's a different topic, but since you know all these AI booming and you know, AWS being one of the major providers, how afraid are you from the bedrock team to generate new stuff and how much are going to be your headache for the next few years?
[00:26:08] Speaker A: Oh, I love it though. I love it because now we're talking about hey, let's build new things, right?
[00:26:13] Speaker B: Like what does the future hold?
[00:26:15] Speaker A: Is it, is it. I mean, I mean the number of opportunities is sort of endless with gen AI and LLMs. But you know, we, we launched a couple weeks ago a general availability of Amazon Q for cost management and like we had to invent new metrics to track the, the accuracy of the responses. Like there was no such thing as a KPI for Cost Explorer query accuracy coming out of a chatbot.
[00:26:45] Speaker B: So we had to invent that metric.
[00:26:47] Speaker A: Actually a, a peer of mine, Liam.
[00:26:49] Speaker B: Built and designed a brand new accuracy tracking concept for us to identify what.
[00:26:55] Speaker A: We felt comfortable with ga and it was really cool.
But not just that. Like we built GPU optimization recommendations in the Compute Optimizer last year. Like it's kind of a brave new world and I'm excited for that because like I had this experience earlier in My career where when cloud was starting to become popular a lot of people I was in it. So there's a huge number of people doing patch management, agent health management, right? So these normal IT practices, there were some people that were afraid, hey, in the cloud I'm, I don't have my patch job, what do I do? And other people are like oh my gosh, that's great.
[00:27:29] Speaker B: Now I don't have to patch server.
[00:27:30] Speaker A: And track patching success across it anymore. Now I can do something more impactful. And if you embrace change and are excited about the capability of, of new technology like LLMs and gen AI, the.
[00:27:45] Speaker B: You can scale yourself and your impact.
[00:27:48] Speaker A: And especially finops, right? If we talk about the culture change and, and helping explain the impact of recommendations and right. Sizing and cost as a, as a service or as a resource to engineers. What if, what if we take LLM capabilities to explain the recommendations and explain.
[00:28:09] Speaker B: This does take into account CPU performance differences and here's the result.
[00:28:13] Speaker A: And what can we do to accelerate finOps with AI? Or again on the flip side using AI to, to do finOps as far as like what are, what are we optimizing in the AI space, right. Are we going to build recommendations for bedrock, right. I don't know yet. Like we're working on it. We don't know what that looks like, right?
The. I get more excited about these developments than I get worried because I, I. If we're, if we're still working on.
[00:28:41] Speaker B: The right sizing problem and getting people to take recommendations five, 10 years from.
[00:28:45] Speaker A: Now, I think that we've probably not done as well as we could have, right.
[00:28:50] Speaker D: AI to advice on AI recommendation.
[00:28:54] Speaker C: It's the, is the endless loop you.
[00:28:56] Speaker A: Can, it's kind of already there though because like our gpu, right Size recommendations, right. We use models to do that. So it sort of is AI for AI a little bit.
[00:29:07] Speaker C: So yeah, it never, it never ends. Like AI for PhenOps, PhenOps for AA and then the cycle will start again. It's, it's super fun.
[00:29:19] Speaker D: Actually. I'm very happy you said about the on the culture movement because it made me think that, you know, you're correct. We need to talk about the culture, about the need.
Because at least me, I'm always, I close it for the better, for the best, for the best performers. I always invite the best.
I guess all the engineers are the same because now they have all these toys, right? If it's like the shop that you open to them and they want to use the better the Best and only these discussions will help make those things different and think about not having the best toy, but the most efficient for the company. From, also from business perspective as well.
[00:30:09] Speaker A: We have to lean in on that, right? Engineers are, I like this one. I've heard this one. A few different folks talking about this, but engineers are lazy.
Engineers are, I would say engineers are efficient.
[00:30:23] Speaker B: Engineers are good at finding the fastest.
[00:30:25] Speaker A: Way to accomplish something. And so if we can help engineers.
[00:30:31] Speaker B: Get to the most performant and efficient.
[00:30:34] Speaker A: Resource configuration types and sort of deliver that to them and make it easy and trustworthy, they'll take it.
And there's a, there's this push and.
[00:30:45] Speaker B: Pull when it comes to optimization where.
[00:30:48] Speaker A: I, I see, it's this cultural thing where sometimes I meet with a company.
[00:30:54] Speaker B: And we show them our recommendations and.
[00:30:55] Speaker A: They'Re like, well, you know, we can't take these and we know our systems and you, and you can't tell us what sizes to use because you don't know our systems. And we do.
I see that response and we need to do better. Building trust. And then I also see the other response where it's like, oh my gosh.
[00:31:18] Speaker B: This is so much more efficient and I love efficiency and it's a big part of how we build great services.
[00:31:23] Speaker A: And there's, there's buy in and how do we convert customers from group A to group B where they're sort of adversarial and they're like, don't tell me how to size my resources. I know what I'm doing. Versus okay, we're on the same team. We have the same goal of efficiency.
[00:31:40] Speaker B: And effectiveness and great engineering.
[00:31:43] Speaker A: And it's not easy. You know, there's some engineering organizations that culturally have commitments to efficiency and doing.
[00:31:54] Speaker B: Things that way and there's some organizations that don't.
[00:31:56] Speaker A: And, and we're all, as a finops.
[00:31:58] Speaker B: Industry, we are all working to do.
[00:32:01] Speaker A: That transition and it requires a huge amount of trust. Building transparency and how we make decisions, how we generate recommendations, providing access to.
[00:32:12] Speaker B: Engineers so they understand the science behind.
[00:32:15] Speaker A: A recommendation, things like that.
Because we really want everybody on that same page where we're all on the same team. We're all like finops collectively as an.
[00:32:26] Speaker B: Industry, as a loose set of organizations and practitioners.
We all want cloud to be effective and high ROI and business valuable.
[00:32:40] Speaker A: We want it to empower and supercharge.
[00:32:42] Speaker B: The next wave of investments that our.
[00:32:44] Speaker A: Companies are making, the next wave of.
[00:32:46] Speaker B: Amazing products and features and capabilities.
[00:32:50] Speaker A: Right.
And we need to make sure our engineers are aware that we want to empower them.
[00:32:57] Speaker B: We're not coming at them with a stick saying, hey, get in line, you're spending too much. We want to come at them with.
[00:33:02] Speaker A: Saying, like, let's make you more effective. Let's come alongside you and stretch your dollar so you can do more and have more impact.
[00:33:11] Speaker B: And that doesn't always come across in finops conversations.
[00:33:15] Speaker A: It's tough, right? And it's a little bit of Corey says this. Corey says it's like marriage counseling for engineering and finance.
I like that quote, but I, I, I feel a little bit more empathy there. I feel like it's more like how can we be the supporting staff for.
[00:33:36] Speaker B: A great engineering team?
[00:33:37] Speaker A: I feel that way myself too.
[00:33:39] Speaker B: My engineering teams that I am counterparts.
[00:33:41] Speaker A: With, how can I make sure they.
[00:33:43] Speaker B: Understand their impact when they're building something efficient or cost effective for their customers? If I can bring some engineers into these customer conversations, I do.
[00:33:51] Speaker A: And they get to see the excitement.
[00:33:54] Speaker B: And the interest on the face of.
[00:33:55] Speaker A: The customer when they see a big.
[00:33:58] Speaker B: Problem area that they're struggling with is now solved.
[00:34:02] Speaker A: And that's, that's important because if you can get engineers excited about their increased impact by thinking about cost, an excited.
[00:34:12] Speaker B: Engineer, an impassioned engineer, is an engineer.
[00:34:15] Speaker A: That builds an inspired product.
[00:34:20] Speaker C: Totally agree. Coming from the engineering background. Totally agree.
Engineers need much more. So now it's more like a mandate or something that we have to ship and engineers have to do everything.
Instead of helping the engineers do their job better, I'm measuring them because of their impact. Instead of, you know, saying that, yeah, you have to, you are wasting too much money. And I was like, no worries, okay. I will tear down everything. No worries. It will be fine for me. Yeah.
[00:34:52] Speaker A: $0 buys everything to a T2 Nano. And yeah, yeah.
[00:34:59] Speaker D: I think we have a new quote. We are on the same team. We need to add it to the real one of the podcast.
Every printout should do that. We got in the same thing.
[00:35:10] Speaker C: Yeah, it's, it's super fun to see like how the trends are getting there and how they talk about CFL and engineering priorities and all this stuff is, is going. Because I think it was a bit too pushy on the. Yeah. Sieve left. Everything safe left. And I was like, how, like do you think engineers know first know how and then they care instead of like having the application running. That has been the metric for several years and it's scalable and on Black Friday works and everything works. And instead of make the cost a non functional requirement that the business is worried about Instead of yeah, this product is amazing, it looks great, blah blah, blah, blah blah.
[00:35:55] Speaker A: I hear that a lot. Where it's like and coming from an.
[00:35:57] Speaker B: IT background the first half of my career it's like if it's not broken, don't touch.
[00:36:01] Speaker A: It hardened to rule and it even goes back to data center days where if I would reboot a physical server and it hadn't been rebooted like, like.
[00:36:10] Speaker B: Turned off and back on in several.
[00:36:12] Speaker A: Years, the capacitors in the power supply.
[00:36:15] Speaker B: Would pop and you lose the server.
[00:36:18] Speaker A: So don't touch it, do not touch it. And if there was a patch that required full power off on we would.
[00:36:26] Speaker B: Have to go send data center staff to go physically watch it to make.
[00:36:30] Speaker A: Sure it turned back on.
So this. And then of course we had ITIL.
[00:36:34] Speaker B: And Six Sigma in the IT world. This is basically a bunch of framework.
[00:36:38] Speaker A: Around don't touch anything, don't break anything. And then here cloud we're saying touch everything, break everything, learn scale, auto heal, resiliency. Such a, such a polar opposite from.
[00:36:53] Speaker B: What the IT industry found is success.
[00:36:55] Speaker A: To keep data centers healthy and running.
[00:36:58] Speaker C: Yeah, it's super interesting how things are evolving on that matter.
[00:37:04] Speaker D: I think that's to the point kind of how it was. I love this conversation with very types.
We would like to ask you what do you now the dwelling is Christmas time. What do you wish for the printout? So what would you like to see next year? What are the thoughts?
[00:37:29] Speaker A: What would I like to see next year? I would love to see FinOps teams.
[00:37:37] Speaker B: Finding that next wave of success with.
[00:37:39] Speaker A: Right sizing and I'll tell you the number of optimization conversations I have with customers, finops teams, engineering teams.
It continues to be one of the top conversations in all of cloud only only sort of passed by AI right now but over the last five years it's generally been and maybe kubernetes before that but it's largely been like people love talking about savings. Love it. It's a fun conversation. The numbers are big. The savings numbers you see in the tools are huge.
We look at metrics, we look at how many people are looking at the savings numbers. It's really big versus the number of people taking the recommendations.
So people love to look at huge savings numbers and talk about it, get excited about it.
[00:38:28] Speaker B: Yet we still see a very large.
[00:38:30] Speaker A: Gulf between the people that look at the savings numbers and the people that take the savings numbers.
And in 2025 I want to continue to see organizations turn those numbers into targets and to shrink the numbers of.
[00:38:52] Speaker B: Potential savings and grow the number of.
[00:38:54] Speaker A: Achieve savings and there's a lot of work we have to do. On the KPI side. There's some companies that have built interesting KPIs there.
[00:39:03] Speaker B: I want to see some standardization of optimization KPIs.
[00:39:07] Speaker A: I've got some ideas for what they could be. I also don't really want to define them. I want to see what customers come up with and define as their exit criteria for optimization. And I really want to see in.
[00:39:17] Speaker B: 2025 the needles moving.
[00:39:19] Speaker A: I've seen the needles moving. There's.
[00:39:21] Speaker B: There's a handful of companies that are extremely mature in finops that have taken.
[00:39:25] Speaker A: A bunch of Right. Sizing recommendations and are very serious about it.
That being said, it still totals a fraction of the potential.
So I want to see that spread. I want to see adoption of recommendations.
[00:39:40] Speaker B: Really become a standard practice for FinOps teams.
[00:39:47] Speaker C: That's a great wish. I love it. I really love it. I think yeah the KPI conversation is going to be sticking for I guess ever.
I think it's the business related stuff. Yeah. Seeing more achievements, see what companies can do. It's going to be amazing in 2025 especially now. Even though I think it will be maybe more in 2026 when they break the budgets with AI in 2025. So then in 2026 they have to apply the finOps.
It will be fun as well in 2025.
[00:40:24] Speaker D: 2025 Garden reports it went from 30% waste to 2%.
[00:40:34] Speaker C: Let's see, let's see how shields.
I highly doubt it. Especially with a.
So Rick, it's been a pleasure to have you here. I think it was a lot of insightful, insightful stuff.
Let's see if we can probably meet next year and Merry Christmas and have a great new year.
[00:40:57] Speaker A: Thank you so much. Thank you Victor. Thank you Damian.
[00:41:00] Speaker D: Thank you for being here with us.
[00:41:04] Speaker C: Happy New Year everybody.
