Kion’s Tatum Tummins on Optimizing Cloud Spending

This is Textron tv. Hey guys, thanks for the throw. We're here with Tatum Tummins, who's a finops product manager for Cion.

And we're talking about, well, what is going on in the cloud and costs because it feels like they're spiraling outta control and we don't have the tools to maybe understand what's going on in there. And even if we have the tools, would we know how to use them? Hey, Tatum, welcome to the show.

Thanks for having me, Mike. Good to be here. So what's your assessment of the current state of things?

I, I, I feel like we basically gave developers free reign and developers do what developers do, and then we're surprised when things get a little expensive. So where is this whole process for cloud cost control breaking down? Yeah, it's interesting.

I, uh, I, I think about how, uh, my entire life, I've had a smartphone, right? But I was an Android user and I recently made the, uh, switch to, to Apple, and, uh, it's taken me months to figure out little things that I would've never expected. And you think about that little case where millions of dollars aren't on the line, right?

When you're talking about a company's cloud infrastructure, I, I think we're setting up our technical folks, uh, to not be successful because we're handing them tools to go build and go create without actually taking a moment to pause and teach them how to do it really well. Um, and in some cases we've got folks who have spent their entire lives working on, on-premise data centers, and the cloud is a brand new technology, uh, and we're not setting them up for success. It feels like a lot of them, shall we say, uh, learn by doing.

And so they jump in there with, uh, you know, APIs and SDKs or whatever else they can find. And, but do they really understand the economics of the cloud and what it is that they're invoking and can they correlate that to a cost? Yeah, I, you know, I, I think the concept is easy for folks to understand that you pay for what you use, but when you're actually in the cloud, uh, making decisions, I use something as simple.

We see it all the time. I saw it as a practitioner in a former life and now talking to customers in a, in a product management role where little things like, I deleted an instance, I expect that cost to, to end, but what they don't realize is, let's say maybe the, the storage volume that was associated with that, uh, instance didn't get deleted, right? That's like an extra step.

And if you've never taken that step before, you have no idea that you just created waste that might sit there for 30 days, 60 days, 90 days before finally someone, or a tool flags it as waste. And so there's countless examples across services on all the cloud providers, uh, of things that until you do, you don't learn. So who's driving the adoption of these finops processes to kind of bring some adult supervision who wakes up in the morning and goes, yeah, we gotta get our arms around this.

Yeah, it's interesting. So, you know, uh, I started in the, the finops space about seven years ago before that term was really popularized. And there were folks like me who just knew nothing about cloud technology, but had a finance or business background and just wanted to help and got involved.

And then about four, five years ago is when the finops Foundation really took off and started to put some terminology and naming to some of these things. Um, and since then, you know, there's now an entire, you know, curriculum set around, Hey, this is what people like you are struggling with, and here's ways to approach your problems. Here's tools that approach your problems and those things.

Um, what I think is interesting is we're seeing companies start to buy in a little bit more on the idea of what finops is. And it's not just trying to save money or cut costs. It's ultimately a way that you, it's an intersection point of how you should run your IT organization.

So it's a way of investing, I would say almost cross-functionally in teaching, even in some cases your technical resources. How to think a little bit more on the cost side. Um, there's now tools, uh, in every kind of niche of the market that are trying to solve and make it easy for engineers to take action on kind of the infrastructure structure that they own.

Um, but it's also folks like myself in our prior life that, uh, just want to help solve the problems, right? Help make things cheaper, help get, you know, grow the business value of the cloud. Um, and so what's interesting is finops can be done by an engineer, a technical resource, or it can be done by, uh, you know, a financial analyst.

It's kind of a catchall As we go through this whole process. Um, it seems like we just need some adult supervision, but where does that get put in place? Because theoretically I could put some sort of control mechanism next to the developer and give them more transparency into what the costs are for what they're consuming, or do I need more guardrails in the DevOps workflow and I'm just gonna do it more from the platform engineering side of the equation?

Yeah, so it's interesting. So right now, the, the state of the finops survey just came out, uh, about a week ago, maybe a week and a half ago now. Um, and in that they list kind of the key priorities and you see enabling automation kind of creeping up the priority list year after year.

And what I think folks are looking for in that, in that realm is kind of what you're referring to is how do I put these guardrails that allow my technical resources to go be successful, go build great products and applications, but have something in place that stops them from these cost overruns. You, you hear these horror stories of folks, you know, losing thousands and hundreds of thousands of dollars in a day because they just weren't aware something was running. And so there's lots of spinoffs tooling out there that just has, uh, simple anomaly detection or enforcements of saying, Hey, if my developer spends over X per day, I want to know about it, or I wanna shut off all capability of continuing to develop.

And so you don't even need to do that in-house anymore. There are tools that come with that capability out of the box. The, the difficulty is, um, I think when people buy a tool, a finops tool, if you will, they think that their problems are solved and they don't realize that they need to learn how to utilize the tools to, to, you know, uh, gain success from them.

So does that mean we need to put everybody back in finops training, which on the face of it sounds a little bit kind of like going to traffic school and maybe as as exciting, I think that, uh, all positions within a company would benefit from some sort of finops curriculum. So, you know, the foundation offers what they call like a practitioner certification. That's kind of the basics of like, what is finops?

What are the things that you need to be considering? What are the processes you need to know? They also recently launched a, uh, engineering finops focused course.

Now, I have not taken that myself. It's rather new. I think we're within maybe two months now of it being launched.

So I can't necessarily speak to its value directly, but they are trying to create content that is both role specific while keeping this idea that finops is a cross-functional, anyone can can implement it at their company concept, Might AI one day just kind of save us from ourselves. We can have a bunch of algorithms figure all this out for us. I mean, we got the data, so, you know, why not just make this one of the first use cases for an ROI for investing in ai?

Yeah, it's interesting. I think we're gonna see that's another one of the priorities you keep seeing, picking up year after year when people are, are experimenting with what AI is gonna do in terms of eliminating cost. I do think one of the things that's tricky about the cloud, and when you're talking about production systems and customers using what it is that's running on the cloud, uh, I think we're gonna continue to still need some sort of, uh, human involvement, uh, for, for a long time here.

Because as good as AI can be, it's still sometimes missing the context of, you know, why does this instance need to be running 24 7? Uh, because there might be a customer that only logs in once a month, but whenever they log in, they need to be able to log in. And sometimes, uh, you know, the, the algorithm itself might be suggesting that it needs to be right sized or even deleted.

And so I think there will be advancements, advancements made, but in terms of full AI automation, um, you know, we're still gonna need people who know this and know how to evaluate, you know, where savings might be. Um, I think of forecasting for example, that's a place that a lot of businesses need help with. What am I gonna spend over the next year, two years, three years?

And they're looking for AI to try to make some suggestions on, you know, this is what my run rate looks like. And once again, it's still missing a lot of that human context of, well, what are your migration plans? You know, know what's your, your revenue plans?

You know, what are some things there that, it's just hard to make that full picture strictly through automation As we get down the path. Um, are the cloud service providers doing enough here to educate people and walk 'em through this thing? Or do they have kind of a vested interest in quote unquote making it as easy as possible?

It's always a funny conundrum, right? They wanna support their customers, but they obviously make money from their customers, right? Uh, I will say in the last year, I feel like they have done, uh, they've made great leaps in terms of just making their platform a little more cost focused in trying to make things easier.

Uh, the four major cloud providers, AWS Azure, Google, and Oracle have all now committed to what's called the focus standard. I don't know if you're familiar with that, but it's basically this initiative to get all cloud billing and eventually SaaS billing, uh, into the same format, right? So everyone's speaking the same language columns are the same, and, uh, they've all made commitments and have already shown like dedication to that.

So they are moving the needle. The problem is, uh, and I don't have, uh, the current 24 stats in front of me, but of most organizations are multi-cloud. And so being able to see all of your data and make decisions and, and, and drive cost savings or efficiency, you need to be able to see all of that in one place, which is where tooling often comes in, is to try to give you that full picture.

And, uh, that's where, you know, there's still gonna be a big gap for a company that's multi-cloud, Our, are people reevaluating what workloads go in the cloud because, um, we have the Covid era and that was the only place to run anything. And I think a lot of people may be looking at that now and saying there are still certain classes of workloads that probably run more cost effectively outside of the cloud. So, uh, in the current economic climate, are you seeing people revisiting their cost analysis for the cloud in general?

Yeah, it's interesting. So I think there was a, in January there was a company that did a a, a big blog post on how they actually reverted completely out of the cloud. It went back to on-prem a hundred percent.

And I've seen some other places discussing that as a strategy, what I've, I, I'm sure there are use cases where, uh, it truly is more cost efficient for some of, uh, for some applications to still not run in the cloud. But I think what's happening is the sticker shock of mismanaging those migrations to the cloud. And kind of going back to this lack of training, right?

Not teaching engineers how to architect effectively, right? For cost or how to watch out for waste and opportunities to make it cheaper once it's in the cloud. I think it causes some of the c-suite executives to, to rethink, right?

What was once their strategy, and I, you know, I struggle and, and this is where I, I wish I could help everyone is kind of going through the analysis of, is this actually more expensive in the cloud or is it just more expensive with how you've set it up in the cloud, right? That's, that's the key. And, and so to answer your question, I do think we're seeing a little bit more, uh, tension, right?

With folks not just defaulting to everything cloud first, uh, but for the most part, I think it's still gonna continue to, to grow and it'll be more edge cases where we see folks, uh, go the other way. So is this roughly the equivalent of I don't know how to drive a Ferrari, so I'm gonna go get my Buick? I mean, Yeah, and you know, the other thing I find interesting is especially with how much like mergers and acquisitions happen, every, you might have a company with a lot of, uh, great knowledge and great skillset within one of the cloud providers, and then they get purchased by a larger company that uses a competitor, right?

A different, different cloud provider. Now you've got technical folks who've spent the last decade getting really good at something, and you're gonna ask them to learn slightly different technology and, and learn how to then take these AppSec and maybe migrate to the cloud provider of choice. And it's even something that subtle, uh, is overlooked where now you're not as familiar, you're, you're, you're essentially building applications that are, uh, less efficient and the costs go up and that's when, you know, regret or, uh, maybe a misunderstanding of what's going on takes place at the leadership level.

Are people moving workloads between cloud service providers because of cost issues these days? There's historically there's tended to be this, we made a decision and that's our primary platform and we've got some sort of, uh, enterprise volume agreement that we're trying to hit. Um, but are people coming back to that now and saying, yeah, maybe certain workloads actually runs significantly less costly on a different clouder?

Yeah, it's definitely happening. You're right there. The, these cloud service providers are very sticky, right?

Once you're there, you're, you're often there for good. But there's a couple things that are happening, and I won't get too specific, but, you know, there are, uh, certain companies that are going into other markets that might cause you to rethink, do I want to use this service provider when I might be a competitor somewhere else in my business, right? So then, uh, you might need to make the shift for competitive reasons, or in my prior life, we had customers who came to us and said, we're actually going only going to use, uh, you know, workloads that exist in this service provider, which caused us to rethink decisions that we'd already made and force that migration from one to the other.

So it's still happening, but there are a lot of tools that have made advancements in showing or trying to show which service provider is best for your application and doing that analysis for you. And I think as those gain traction, we're going to have folks who realize I actually can make a shift to a, to another service provider and have a cheaper, you know, workload for the next 10 years if I'm willing to make the, the investment for the next, you know, couple months of making the migration. We are also seeing a lot more workloads that are AI driven.

Is is, does that change the math on the cloud? Because now only do I have to train it, but now I got the inference engine and I'm not quite clear where that always goes. And GPUs are hard to find and they're expensive and there's all these alternative processors.

Is it gonna get harder? Yeah, I think, uh, before I, I changed roles. We were really, you know, making a lot of, uh, investments in innovation in, in that area, and we definitely had some horror stories with, with cost and, and you know, the, there's a couple instances where, you know, overnight or over on the weekends something was left running and it was, it was doing its machine learning, right?

And that had cost associated with it. I think this is an area that folks, uh, are not necessarily prepared yet to handle on like what the cost might look like. I also think it's an area though that companies know is a little bit uncharted, and so there's like perhaps more of an expectation for cost volatility.

Um, but we're definitely gonna see advancements in the tooling side trying to help companies get their arms around how do you control, you know, these machine learning innovation bets that you're making, um, from a workload standpoint. All right. So what's your best advice to folks that are trying to figure out how to wrap their arms around finops?

My best advice is find the people in your organization that are either passionate about this or have the skillset to start learning and building out some sort of finops program. You know, I think the mistake that companies make is they think they need to go invest in a, in a resource that doesn't exist the company, right? They need to go pay someone a headcount.

And I don't think that's necessarily the case that most companies have folks who have the skillset and the knowledge to start to disseminate, you know, internally the education needed for, here's what you look out for, here are some processes we need and start to build out what is typically referred to as like a champions program. You need 10 ops champions within your org that are, you know, cost conscious and thinking about these things. Um, from there, many, many companies need a tool or some sort of third party support.

You know, I think often those, the cost around those, it's, it feels like it's, uh, just worth our investment. But I think what companies will find is, as your cloud cost pro, you need that extra support of having automation built into your, to your cost controls. And also just give your organization a single pane of blast of here's the visibility into what we're spending each day.

Uh, those things go a long way. And so go back to your question, my piece of advice is find the folks in your organization that are passionate about this and empower them to start kind of building out a finops practice. Alright, folks, you heard it here.

There was a time when we tore our hair out over simple things like capacity planning and on-premises environments. And today the cloud makes all that look like child's play, but that which you do not pay attention to gets expensive really quickly. Hey, Tatum, thanks for being on the show.

Yeah, thanks for having me. All right, I'm back to you guys in the studio.