FinOps in Action

FIA - Cristian Magherusan-Stanciu
===

Cristian Magherusan Stanciu: [00:00:00] nothing really moves the, the needle until you take action. So a lot of companies just get started and build all, sorts of fancy dashboards. And then not nobody takes action, nothing happens. And then like, yeah, they, wonder like, why is the cost, uh, why is inefficiency still there? Why isn't the cost dropping?

Intro: Welcome to FinOps in Action. I'm your host, Taylor Hauck. Each week I'll sit down with FinOps experts to explore the toughest challenges between FinOps and engineering. This show is brought to you by 0.5, empowering teams to optimize cloud costs with deep detection and remediation tools that actually drive action.

Taylor Houck: Hello and welcome to another episode of FinOps in Action. Today's guest has seen cloud cost optimization from almost every angle. He started his career over 20 years ago in systems engineering and spent time at companies like Nokia and HERE technologies before joining AWS. As a senior [00:01:00] specialist solutions architect focused on EC2 SPOT and graviton in 2016, he built auto spotting an open source cloud cost optimization tool that has helped more than 10,000 customers save money on EC2.

Then he went and built something of his own. Today he is the founder and cloud optimization specialist at Leaner Cloud, where he works hands-on with midsize companies to optimize their cloud environments. Cristian, welcome to the show.

Cristian Magherusan Stanciu: Thank you, Taylor. Thanks for having me.

Taylor Houck: Absolutely. I actually was really looking forward to this episode, Cristian, because I think that you have. About as much hands-on experience optimizing cloud environments as anyone in the world. What is the number one thing that you wish that every organization knew about FinOps and cloud optimization?

Cristian Magherusan Stanciu: Yeah, I mean, it's not that that, people need to know, but like, uh, nothing really moves the, the [00:02:00] needle until you take action. So a lot of companies just get started and build all, all sorts of fancy dashboards. And then not nobody takes action, nothing happens. And then like, yeah, they, they wonder like, why, why isn't, why is the cost, uh, why is inefficiency still there?

Why isn't the cost dropping? yeah. So I think it, it's very important to be hands on and Taking action as much as possible on, on the insights that you get from all of these dashboards. I'm not saying they're wrong, it's just that it's just half of the story. Uh, people have to also like pay attention to them and, and, uh, yeah. Do the, do the work that needs to be done as a, as a result of those.

Taylor Houck: It, it's

Cristian Magherusan Stanciu: course. There's always, uh, a lot of the things can be automated, so you can, you can always like. Have, have some help from the automation that exists in this space. as you said, I've been building tooling for, for this since [00:03:00] very early on. Um, and I'm, I'm a big fan of the automation, but then again, automation can only go so far. Um, there's always. Some, some gaps or, or things that, that you need to do, like taking over , from where automation leaves you. And then like, kind of what I do is kind of bridging this gap between what can automation do and what actually needs to be done to cover everything properly.

So yeah, that's, that's basically what I do nowadays.

Taylor Houck: I think that's such an important insight, right? Because so much of, You know, the FinOps industry is centered around cost visibility, cost reporting, cost allocation, but what do you actually do with those insights? That's what's important. That's what actually changes something. You can look at the data all you want, but if nothing changes, what's the point?

How do you drive action? How do you ensure that engineering teams are able to actually act on the insights that FinOps brings to [00:04:00] them?

Cristian Magherusan Stanciu: my specialty is more into the, into the companies that basically don't even, like, they, they don't have anything at the point where, where they involve me in their, in their setup. So most of the times they just have the default, dashboards from like cost Explorer. And basically I take them from, from that level, from level zero. And I, I go hand in hand with them and like, take actions, like looking to the things that can be done and drive the entire implementation, um, together with them, but at the same time. I do it as a service, so I, I kind of try to offload as much as possible of the gory details, if I can call it like this, um, the boring things that maybe it's not like the best use of, of, uh, of their time. Um, like. Things like, like just optimizing GGP two volumes, like those, those, those tiny things that maybe doesn't wash their time and, [00:05:00] and, uh, and attention. I can take care of them with automation, and like making sure that everything is, is implemented in the right way. Yeah, there's, there's always, uh, like process that they need to follow. So like maybe infrastructure code, avoiding drift and, and stuff like that, which is a big topic of what I'm actually working these days. yeah, so like rolling out changes in, in a lot of, environments. Like they have hundreds of, of databases that need to be right sized or, or, um, Converted from, from X 86 to Graviton, for example. yeah, and like all sorts of, of misconfiguration or inefficiencies that could move a bit, the needle that in, in the end, they add up to significant amounts of money. But then individually, each one of them maybe not, not worth it for, for them to [00:06:00] do themselves, or they're just maybe too busy or don't know all the details that, that, um, they could take, they could take action on.

Taylor Houck: Engineering bandwidth is oftentimes the most scarce resource that we have.

Cristian Magherusan Stanciu: Yeah.

Taylor Houck: What do you think is the right way to work with engineers to make sure that they're paying attention to. The highest and best use optimization opportunities that are at their disposal and not getting distracted by perhaps the the small things that could be optimized.

Cristian Magherusan Stanciu: so there's, there's always like prioritization that needs to be done. You cannot just go to the engineer with like a $5, opportunity and expect them to take action. They will always say, yeah, this makes no sense. Um, like. Come to me like next year or when you have something bigger. But at the same time, there are things that maybe fall through the cracks that are, are more sizable. [00:07:00] And, uh, yeah. So I think it's very important for, um, for the teams to have in place all these insights and dashboards that give them like an idea of, of where they stand and, um, what could they do. the, the most impactful things they could do. Um, and then actually having the bandwidth and the, the allocated time to, to work on those insights and, and drive some results.

Because just as I said, just knowing them as doing nothing doesn't, doesn't help.

Taylor Houck: I want to drill into the topic of automation a little bit because you've already mentioned it a few times. How do you think about automation as it pertains to remediating cloud inefficiencies? Because I'll, I'll tell you something that I've heard and and thought about in the past, and it's the fact that.

When you are taking an action to change a cloud resource, whether it's a configuration change, whether it's actually resizing an instance or something, there [00:08:00] is some inherent risk to the performance of an application when you go in and and change something. On the infrastructure side, how do you think about automation and ensuring that automated actions don't introduce unnecessary risk?

Cristian Magherusan Stanciu: it happened before, like, even, even as part of my work, it happened that, um, maybe we went a bit too far. So, but in the end, um, it's, it's. We should see this as a, as an ongoing iterative process. Sometimes you, you, you notice something, you say, okay, this is completely oversized, and then you, you configure it maybe too small and, uh, yeah, that can, that can uncover problems.

And there, there's like always a risk of, of doing too, too aggressive. Um, I, I think what what helps also like. Like taking it step by step. Like, not if you have something that's, that's, um, running at 2% utilization, you don't rightsize it all the way down. Maybe you, [00:09:00] you cut it in half at first and then like let it run like that for a while and then maybe see if, okay, how, how does it go afterwards? Maybe cut it again in half. But not like go all the way, like cut it in, in, uh, in like eight times from a single step. I mean, it's not rocket science. There's always like, like these low hanging fruits, like right sizings and like small misconfigurations or, or suboptimal configurations. but yeah, it's, it's like being a bit cautious and error airing in the. Um, how can I say, towards caution rather than, than over, uh, over optimizing from the first attempt. Um, that's why, kind of like, I like to, to work with customers for a longer time, that that gives us more time to okay, like, iterate and, and come up with something that actually makes sense. Not just tell them to cut their instance. In eight or [00:10:00] whatever. So like, just see how, how much can we, can we squeeze it? But without introducing too much risk.

Taylor Houck: Talk to me about the, the people aspect, right? Because I imagine that when you come in and you're working with the. You're gonna be working closely with, You know, many of the folks on their engineering team, who are the personas that you're working most closely with, and how do you collaborate with them in a way where You know you are taking their perspective and specific knowledge about the actual applications that You know are running in the cloud, and ensure that that is taken into consideration as you propose optimization opportunities.

Cristian Magherusan Stanciu: um, engagements I had were with, uh, in like people from the engineering team, maybe DevOps engineers, platform engineers. Um. Even like senior developers that, that know very well their application. so basically we, like, when I come up with something, I, I always like [00:11:00] discuss it with them. Like, does this even make sense for you guys to do? Uh, of course there are some like, like low hanging fruits that maybe don't warrant much of a conversation and we just say, yeah, we need to do this, and then we do it. there are things, uh, sometimes that that needs to be discussed. Okay. um, a recent example was, was converting some databases from, um, Aurora.

I optimized back to Aurora Standard. And uh, yeah, there were like cases that added up to a few thousand dollars a month, but they were like more a conscious decision in a way that. They, they try to like it. They see it more, more, um, as of a sort of insurance a little more for, for Aurora io optimized even where it's not, not necessarily needed. Um, but like, yeah, like, um, discussing this, this kind of aspects, like explaining [00:12:00] all the, the pros and cons of the options seeing what, what actually makes sense for them.

Taylor Houck: You know, when you think back on your experience, have you ever worked with a client where you had a proposal that you are, let's say, very confident in, right? But perhaps an engineer or someone within their organization, um, pushes back on that due to risk. How do you handle those types of conversations and what's the balance?

Cristian Magherusan Stanciu: a recent example for that was, uh, was when we, uh, were trying to optimize some, some EBS volumes. Um, so we had a bunch of IO two or IO one, uh, volumes that we wanted to propose converting to GP two. GP three. Um, and when looking at the metrics, like the volumes were pretty idle like most of the time.

And then occasionally they had some, some IO going on. Um, they had always some like very tiny but there was [00:13:00] like very, very little load. But the end it turned out didn't want that because of latency concerns. So because like GP three. Has increased latency and the, the, the application itself, um, is very latency sensitive.

So we opted to keep, to keep them as, as, um, io uh, IO one, instead of converting to to GP three. So like there's also all these, all these kinds of insights. That maybe from outside looks like you could do them, but sometimes, yeah, we realize it's not really a good option. then what we, we can do is like making sure that we configure the, IOPS and throughput as much as like is actually needed, like right sizing within that particular kind of configuration. Um, because it was initially kind of severely oversized also from that perspective. [00:14:00] Um, so that's, that was how we eventually like, brought the cost down, but um, not being able to go all the way to the, to the like more, more optimal configuration the way we saw it initially because of these application latency concerns.

Taylor Houck: I think it's important for the listeners to really digest and understand this because. Oftentimes people that are solely focused on FinOps. You can call it FinOps, you can call it cost optimization. People where cost is the primary metric that they care about and are thinking about when they're looking at a cloud environment.

They're prone, especially earlier on in their journeys to dismiss or not even consider alternative uh uh, metrics or alternative. Benefits of being on a certain, You know, configuration or resource or architecture. I mean, you used the word earlier insurance, [00:15:00] right? Oftentimes it, it, it can be. A bit of an insurance policy to be overprovision.

Sometimes you have HA or DR scenarios where you need infrastructure that appears idle from the outside looking in, but it actually is critical to have that in place in a situation where you need to fall back on that. You need it to be there depending on what the application is. So it's really important to, to take that, um, perspective into consideration and be open to other perspectives.

Even if it appears that there are easy cost savings to be achieved by taking an action,

but that being said. There are many, many, many times where you can take an action and you will reduce cost significantly, and there is absolutely no impact to performance or availability or scalability or security.

And in these cases, that's like what we're here for, right? It is like, how do I find somewhere to just [00:16:00] take an action and, and save a million dollars annualized? Save $250,000 annualized save. $10 million annualized, all depending on the scale of the environment. Cristian, I wanna, I, I want to touch on this though, and, and, and kind of look back on, on your experience and all the environments that you worked with.

What are the, the, the scenarios and the cost savings wins, um, that stick out to you when you look back? And are there any that have come up Time and time again?

Cristian Magherusan Stanciu: touching back on what you said with like you can do it and like you can get lower costs, sometimes you can get lower costs while actually also improving performance. Which is like a double, double win. I mean, there were cases like, I think the, um, yeah, the, the best that comes to mind was not so far, not so long ago, there was a, a, a database configuration that was like serverless aurora. And basically [00:17:00] like if, You know, if You know about serverless Aurora, you can like see how um, the costs are are like about three, three and a half times more than than provision Aurora for the same size. And it's a very useful configuration if your workload is like very bumpy. And when it comes to traffic. Um, like in case you are, you're running idle for like 99% of the time, but then there's this time of the day or when something comes and you, you are flooded with a lot of requests and, uh, it, it scales very quickly to serve such peaks or, uh, or bursts in traffic. But if your traffic is completely flat. Uh, and you're using, using it, you are paying that, that premium for, um, for no benefits. we've seen such a use case for, for for, uh, for one of [00:18:00] my, my clients and we converted it overnight, like to, you can just quickly convert it to a provision database. It just needs a sim, a simple failover. And what we noticed afterwards was that actually latency got significantly better. Um, because this provision database has like, like maybe a, a better, uh, better tenancy guarantee than, than the, uh, serverless equivalent. It could also be that they run on, on more recent, uh, instance types, so, so faster, hard hardware. Um, but like what we've got was. Um, like costs decrease of like three, three times. 'cause we kept kind of within the same, uh, capacity, size as we were getting before, but also we got performance, um, by like 30% better than before just [00:19:00] by switching to to this.

Taylor Houck: That's an incredible story And something that. It's important to recognize that there are opportunities where you can actually reduce costs and improve performance. And reducing cost by a factor of three is, that's a significant improvement, right? When you think about the underlying, um, You know, application and what it's doing, you're essentially making it much more productive asset.

And the ROI of hosting our database has increased greatly it has a significant business impact when you're able to make a change like that.

Cristian Magherusan Stanciu: Yeah, and this was like something that was costing like 12 k or so a month. uh, initially like, and in the end we, we got it to about 4K and with faster performance. So it, it almost saved the cost of an engineer. I mean, they don't, they don't, uh, they have a bit engineers in, in more lower cost areas, so it, it, uh, yeah, it was pretty, impactful for, for the customer. For just a couple of [00:20:00] changes.

Taylor Houck: Are, are there any waste patterns or inefficiencies that you see come up over and over and over again?

Cristian Magherusan Stanciu: Yeah. I mean, it's, it's always the usual suspects, like oversized capacity, idle resources, um, suboptimal configurations for like storage volumes. These Aurora, I optimized the Aurora serverless where it's not actually needed or not a good fit. Um, the, yeah, like even support costs for, for databases, if we can get them upgraded to the, to the recent versions. So it's, it's not really like, not rocket science, but there's always these things that they are, many of them are, are tiny in, uh, absolute value, but they add up to significant, uh, dollar figures over the entire, estate of the customer.

Taylor Houck: It is funny because I think that a lot of what we do, You know, I've, I've done this type of work for [00:21:00] many years, is you look at the resources, you find the top spend resources, you look at their configurations, you look at the utilization of these different, You know, services and try to understand, hey, is there an opportunity here?

How can. Maybe use the utilization metrics to determine the optimal configuration. But in our prep call, Cristian, you were mentioning that sometimes it's not always what it seems. Uh, I think there was a, a story with, with Aurora that you told me where that you had a particular insight based on utilization, but under the covers there was a, a different scenario.

Can you maybe tell us about, about that story and what the result of it was?

Cristian Magherusan Stanciu: It was about a year ago when I, when I first, uh, get, got started with the client. And the first thing I did was looking at their bill and see the, the, like the top spenders and RDS was always, as, as pretty as, as most of my clients, RDS team tends to be one of the biggest. And within their [00:22:00] RDS I noticed massive costs coming from, from uh, io, Aurora IO costs. I traced them down to a single database. Basically it was costing them about $30K a month, um, for a single database. Just the IO costs, whereas the entire database was like the infrastructure, the, instance itself was about maybe $1,000 or $2,000. So the, IO was like, more than an order of magnitude, bigger than, the compute and storage for that. for that database. and it turned out after I flagged it to the engineers to see what's going on, how much, because we couldn't actually, change it to, I optimize since it was running an old database engine. So the conversation turned more into, okay, how can we get this database updated to [00:23:00] the latest version of the engine? And then one thing led to the other and we realized that that database was actually. Not really needed anymore. So, so the customer were, was, uh, was paying for something that one of their clients kept using, but that client was like, discontinued or, or, or something like that. Like they, they weren't supposed to be using it anymore. Um, so in, in like just a few, a few weeks, we, we managed to shut it down. So we, we got like. 30 plus K a month in, in savings. Just from, from like, like turning it off.

Taylor Houck: I think of this story, Cristian, it, it really paints the picture of, of whether you want to call it FinOps or cloud optimization or this, this role that, that we play in looking at, You know, cloud environments and understanding the cost impacts. It's not all just within the engineering organization.

There's a [00:24:00] broader business context that's really important to bring into the picture and that's where you can move from, You know, hey, let's change the configuration and save 30%. Or even more than 30% to actually, the underlying reason this database is running is not valid and we can just get rid of it.

Cristian Magherusan Stanciu: Yeah,

Taylor Houck: I want to, I wanna shift gears a little bit with you, Cristian, and talk about, um, another topic which is realizing and showing the value of the work that we do. One of the benefits that we have in FinOps in cloud optimization is. The benefits, the value is, is very real. It's tangible, right? You can track it to a realized cost saving, but oftentimes you can perform an action that reduces cost on a specific workload, but the overall cloud bill could still be growing forever.

A number of reasons. [00:25:00] So tracking the value of this work is, is difficult. How do you think about quantifying the value of cloud cost optimization efforts?

Cristian Magherusan Stanciu: I mean for me it's, it's particularly important to, to quantify this kind of impact because the way I charge my clients is based on, on savings. So I, I, like, I charge a fraction of the savings that we agree on, and if I don't deliver savings, basically I don't make money. Um, so it's in my, it's my, my incentive to like track the savings accurately. And like try to Yeah. Um, come up with, with like ways that, that actually make sense that we can deliver like meaningful results. Um, it was like I had a, a case at some point in the past where I was like explaining how I do it. Like I, I just generate savings for, for your infrastructure basically. In plain, in plain [00:26:00] English. And we started to work and then within the first few months. We, we got, uh, yeah, like significant savings in like, the range of 30% of their, of their footprint. Um, but the thing is that this client had been a kind of explosive gr growth for the last year or so. They, their, costs have been growing like almost double in 12 months. So the, the growth was, was continuing throughout my, my time with them basically they just couldn't see how like. The results that I'm driving and the, the, like, the figures that I'm, that I'm, um, I'm helping optimize didn't materialize entirely in their, uh, in their bill 'cause their bills kept growing and like, there was always like a bit less savings, let's say in in absolute number. to how, what, what we were, we [00:27:00] were expect, I mean, they were expecting more. Um, and the thing was like, even after I finished my work, uh, it, it was kind of a bad coincidence that it, my, my backlog was like, okay, I finished the work and then, soon after they had, uh, like Black Friday, which increased the traffic a lot. So their, their costs exploded. They also at the same time had been going into re-architecture of the application, converting from EC2 to Fargate. And, uh, if in case you're not familiar, Fargate, like has this, uh, extra cost on, on logging compared to EC2. In EC2, they were just logging to the disk. So there was no cost. But as they were migrating to Fargate kind of. In at the same time with this Black Friday event, they had a lot of [00:28:00] costs coming. Right after I finished my work, they switched to Fargate and had Black Friday and their costs went way over what I managed to achieve in savings. Um, so we had a, like an interesting conversation afterwards and like I was like, okay, yeah, uh, we. We did, we got some results, but then, then things, things went completely over. Um, yeah, that was an experience way I, the way I, um, started to, to make like it clear afterwards that like, it's, it's not necessarily absolute savings, but it's like cost avoidance compared to how much you would be paying if you'd take no action. Because everything that I, I did was still generating savings for them. Like, they didn't revert any of the work, but like they just Yeah. Had, had more traffic and then like a different architecture for, for some application that, exploded the [00:29:00] cost.

Taylor Houck: look, if you're just running a, a cloud environment and things are very static and you go and you do cloud optimization efforts, you're gonna see your total bill go down, of course. But on the other hand, if you are growing, if you're seeing increases in traffic, if you're deploying new applications,

even if you are optimizing your existing infrastructure, you could still see your total cost go up. When I was working at NBCUniversal, we called it self-funding growth. And essentially what we would do is we would, You know, we were a big enterprise. We had a, a, a very large cloud budget, right? And the budget was set at the beginning of the year.

This is how much, You know, we have allocated in your organization's budget to spend on the cloud, right? Pretty simple now. Throughout the year, we would always have an influx of, You know, there's essentially demand for more cloud services to be used, right? But we only had the budget that was allocated to us.

So what do we do? We embark on cost optimization efforts on the existing infrastructure, [00:30:00] reduce that spend, and use those savings to reinvest back into. These new projects. Now, it's interesting though because as an outsider, right, imagine you are an executive or you are a finance person that is not in the weeds and you're hearing like, oh, apparently these guys are doing all this work and optimization and have saved so many millions of dollars, but now it's the end of the year and the bill is the same.

So obviously, You know, they didn't, they didn't do anything that important, but you're, you're missing the piece, which is what you get out of the cloud is value. Right. And you, when you are doing more in spending the same, it's the same thing as as reducing the the spend, right? Essentially, if, if, if these actions hadn't taken place exactly easy, just described, then the, the cost would've been much higher.

It's not that, that we're not still getting the benefit from the optimization, it's just [00:31:00] that we are also doing more. That then led to the cost being the same or oftentimes, uh, even higher, potentially.

Cristian Magherusan Stanciu: I think it, it boils down to like the concept of, of, uh, unit economics, which at the time, I, I knew about it, but didn't really feel like, okay, this is how I should describe this situation. But yeah, it, it is, um, a way to measure the efficiency. Like how, like like how you can, um, you can get more from the same, from the same spend. Um, yeah, like you, you described, you, you invest, you invest, you can do more investments for the same amount of money.

Taylor Houck: Absolutely. Now, now Cristian, I, I do want to, um, shift gears slightly as we move towards the, the end of the episode and talk about ai. I think it's a, a very hot topic right now and I feel [00:32:00] that a lot of us are. Feeling that the capabilities of these models are exploding and making us much more productive.

I'm super curious from your perspective, how you're seeing AI shaping the, the capabilities of you as an individual and also how you think about AI spend and optimizing those types of workloads.

Cristian Magherusan Stanciu: mean, when it comes to using the AI for productivity increases, um, when I went on my own, like, think it was a few months that, that, um, Chad, GPT-3 0.5, or was it 3.5? The first version, like it was released in 23 at around Christmas. And I got very early on started to use it to generate code, but basically from the browser, just copy pasting from the browser to the IDE, like the, this, I had this, [00:33:00] uh, manual workflow for a while, but I basically stopped writing code for all my tooling development.

Soon after I got that, um, that first version of Chad GPT, and. It was kind of working for relatively small files. So if you could have like maybe a file for around 200, 300 lines, it could handle it and do refactorings inside it, and then like pasting the entire file and then I copy paste it from the, uh, browser to the id. And then I was kind of in this loop of always like looking for, for, um, ways to improve my work through AI like went. And, uh, and kept, uh, kept kind of workflow, but the models kept getting better and better within the browser until last summer where like I had built even some tooling that can take the entire project, [00:34:00] copy it into clipboard with each file separated by file names, and the model was able to like, take that input and like refactorings and just. Give back certain files and I would just copy certain files to the, to the IDE. So I've, I've been doing this for like maybe a year or two years until I got my hands first on this, uh, cloud code. And the first time I used cloud code, I, I was disappointed 'cause it was pretty aggressive in like, just reading like the first 50 lines of the files. So it was not really meaningfully, uh. Like it couldn't meaningfully understand what's going on in that file.

Taylor Houck: When, when was this?

Cristian Magherusan Stanciu: maybe June last year

Taylor Houck: June of 2025.

Cristian Magherusan Stanciu: yeah, so I got back to my, to my browser where you could just paste 10,000 lines of code. Into the browser, into the cloud window and [00:35:00] get entire 500 lines or thousand lines files with as, as, as per my request. then I, I went back to that for a few more months and then I got back in September or October, I gave it another try with cloud, um, cloud code and then it just worked way, way better. And then I kept working on like improving my workflow and like have been using it a lot, especially with the last model 4.6. in November I got on that like within a few weeks basically I built. In auto spotting because I keep working on auto spotting even like 10 years now. It's, it's been 10 years around this time since I started the project and first talked about it online. And then in that like in a few weeks, basically I got to build [00:36:00] that previously I was building maybe in a year or two combined. Um, so now I'm like, I'm very close to releasing the, the next version that will have like a web interface with account support and like, it's even gonna work as a SaaS. So I'm building like a, a like, um, a sa of sorts, uh, for, for auto spotting. Whereas previously it was just running with a, within a single account. yeah, basically I, I, I used the AI for that and also like. Built all the, all the time. I built new tools. like 30 plus tools for various optimizations. around reinvent, I released an open source tool for RI and savings plan, purchase automation based on the recommendations from the, from the cost, um, And within the last few weeks, basically I built a, a web interface for that. it's gonna work more like, uh. [00:37:00] Fully automated purchase engine based on recommendations from the, from AWS cost explorer, um, like stacking the purchases over time and renewing the thing like the purchases automatically, so you don't have to worry. so this is kind of the, like the work that I've been doing for, for like open source and like product work, but also like always building tools for my customer engagements. Like just recently, like this week I've been working on an Aurora IO optimized configuration, uh, drift avoidance optimization.

I, I, yeah. So basically you configured the I optimized flag. In a way that avoids, uh, called Drift. Um, completely automated so that next time you run Terraform, basically doesn't see any drift. [00:38:00] Um, which is, which is what I've been working on lately. Uh, so there's like a lot of benefit when it comes to the, like being how much, how much I'm able as a single guy, a solo. Uh, agency or solo, like software vendor of sorts? Uh, how much I can achieve. It's, it's massive, uh, compared to what I've been able to do when I first got started like three and a half years ago. Um, yeah, so it's like at least an order of magnitude more output per unit of time, if I can quantify it like that.

Taylor Houck: it's absolutely amazing and mind blowing. I mean, I've been using these, these tools as well and had a similar. Kind of, um, realization as you, I mean, I almost feel that in today's day and age, and also if you just assume any rate of improvement, which I think is inevitable, it's just your mind can get blown with thinking of where it's going.

But it's, it's like an individual [00:39:00] has the, the abilities of a team. And then a team has the abilities of like a department, right? And it, it, it's, it's pretty mind blowing and it's gonna be really interesting Cristian to see, um, where, where things go and how it's gonna impact our industry. We didn't even get a chance to talk about the, the growing amount of AI costs that are coming.

Um, but we'll have to save that one for, for next time. Cristian, this has been such a fantastic episode, um, to our listeners, where, where can they find you?

Cristian Magherusan Stanciu: Yeah, I think the best is on on LinkedIn. I mean, I'm, I'm there pretty active, but also I have my website, LeanerCloud. So in case somebody wants to check out my, I have also like, uh, um, a YouTube channel where I was posting in the past some, some, optimization, let's say focused posts and yeah, podcasts and stuff like that. I'm not really maintaining those anymore, but, but it's still there.

leanercloud.com is the entry point.

Taylor Houck: Awesome everyone, uh, uh, check that out. Cristian, this [00:40:00] has been fantastic. Thank you so much for coming on the show, man.

Cristian Magherusan Stanciu: Thanks for having me, Taylor. All the best.

Taylor Houck: Absolutely. And thank you to our audience. If you've got something out of today's conversation, please share this episode with someone that needs to hear it. This has been another amazing episode of FinOps in Action.

We'll see you next time.

Outro: That wraps up another episode of Fit Ops in Action. Thank you for joining. For show notes and more, please visit fit ops in action.com. This show is brought to you by 0.5, empowering teams to optimize cloud costs with deep detection remediation tools that actually drive action.