[00:00:00] John Kundtz: Hi everybody. My name is John Kundtz, and welcome to another edition of the Cloud Collective Podcast. Today, we will be talking to FinOps certified practitioner. And patent holder Srini, as he shares valuable advice for those embarking on a corporate FinOps journey, especially some of the tagging pitfalls and mistakes many FinOps practitioners can make before they ever get started.
Welcome to the show, Srini.
[00:00:32] Sreeni Maheshwaram: Thank you, John. How are you doing today?
[00:00:34] John Kundtz: I'm doing great. It's great to have you on the show. This is actually a topic that I ran into a bunch of years ago when we were trying to do some cloud management engagements, and we found that the client wanted actually to move forward but struggled, and it all came down to they just didn't really have a good tagging strategy.
So, I'm really excited to talk about that today.
Before we get started, You have an impressive LinkedIn profile. Just give us a little more background and how you got into what I call the FinOps business and the tagging business, if you will.
[00:01:14] Sreeni Maheshwaram: Yeah. Thanks for asking. Let me introduce myself quickly. my name is Srini Vasulu Maheswaram. I go by Srini. Currently, I'm working as a principal architect in Kyndryl's cloud practice delivery organization.
I'm also a certified finance practitioner. I have more than 23 years of IT experience. I worked in various architect capacities, such as enterprise architect, principal architect, and solution architect, across North America, Asia Pacific, and EU, across different domains like transport, retail, telecom, and many of the public sector clients.
My primary job is to provide architecture solutions and design complex distributed systems and mission-critical systems. So in my current job, I am actually delivering modern IT operations solutions along with the modern. Operations applications as well. So, cost and asset management is one of the applications or solutions I delivered.
It basically used by many clients for cost management. And that's my association with the tagging automatically. It led me to it because anything to measure requires an association, for whom, and what the metric that we are measuring is.
In the cost and asset management work, very quickly, I realized after deploying it to several customers that the customer stack coverage on the cloud providers or even on the on-prem, typically, is like 30%, not more than that in the initial days, what we have seen. And though the application has powerful features. Because there is a lack of tagging, the outcomes are not so great.
So the value of the solution was not benefited by the customer. So very quickly, I realized there is a need to put together a model, a lifecycle management model for tag management, right? From defining a schema, defining the library and the taxonomy, doing the tag analysis, running a compliance report, doing a reconciliation, and letting the teams do the remediation of the tagging, right?
This is how I got into the tagging. And now, with Kyndryl, it's even more relevant.
[00:03:20] John Kundtz: Let's unpack that a little bit. You talked about what tagging is, but what is it concerning FinOps, and what are some of the benefits companies get once they have integrated tagging into their, I'll call it, their operations or their FinOps practice?
[00:03:35] Sreeni Maheshwaram: That's a great question again. So let me quickly introduce FinOps, or Cloud Financial Operations as a new financial operating model for the cloud. It includes public or private cloud as well, so it basically brings financial accountability and creates a cost-conscious culture by bridging the gap between IT finance teams, the operations teams, the business teams, and the technology teams.
FinOps has three main phases: the informed phase, the optimized phase, and the operating phase.
And in the informed phase, the essence is to bring visibility. Into the cloud, who is consuming what? With the cloud, there's no need to go through a lot of preapproval. Anybody can just simply spin up an instance, and the building starts.
So the central team, like finance, will get bills from various providers, and now it is difficult for them to identify who consumed it and internally transfer that. They will split off that bill to that, so charged by the job purposes. And that's the visibility approach. That's the informed phase that essentially solves this allocation problem when it comes to optimizing in any cloud. There are a lot of opportunities to optimize the cost, right?
Rate optimization, as well as usage optimization. Rate optimization is if I'm consuming continuously over a period of time, a specific instance type of an instance from your provider, I might consider a volume discount option that the provider gives me in terms of reserved instances or the saving plan.
So, for the same on-demand cost, I can probably get a good 30, 40, 60, and sometimes 70 percent discount as well. I can even start using a bring-your-own license and that kind of approach.
When it comes to usage optimization, it's all about do I really have to require that particular size or did I over provision it if my usage patterns are telling me that over a period of time, my CPU and memory usage is consistently being like less than 10 percent less than 15 percent I might have over-provisioned it by two notches. Then, one size to the immediate next size in a cloud is almost double the price.
So, by just cutting down to one size, I can save more than 50 percent. So that's the Operate Phase and the Optimize Phase. In the Operate Phase, it's basically defining a baseline of all my consumption and attaching KPIs, which account is using what over a period of time, whether the trend is increasing by what percent, and whether the business is growing by X percent.
So, tracking all these metrics. As you need economics and reporting. I think that's operations; in order to do any of this what I need to identify a set of resources or a set of cloud assets that are incurring costs, which is getting a bill from the provider. I need to be able to attach it to a cost center.
So that's what tagging is about in the FinOps world,
[00:06:52] John Kundtz: that's good insight. That reminds me of the early days of FinOps, maybe five or six years ago, before it was even probably called FinOps. We were working with a client and sort of what you described: they started bringing in more and more cloud instances, both server and storage and network. And then, at the end of the month, they got a bill, and it would take them anywhere from 35 to 40 days to reconcile who was using it. Was it authorized? What purchase order should it be billed against? And it was taking an inordinate amount of time.
And once we showed them how, once they properly tagged their instances, it took that reconciliation from 45 days down to a week.
[00:07:36] Sreeni Maheshwaram: The cloud It's very dynamic. So as the reconciliation takes time and is restorative by that time, organizations would have spent much, much more already, right?
Getting a handle around it up front and controlling it with a metrics-driven approach is the key. And I think tagging helps, as you mentioned.
[00:07:57] John Kundtz: It reminds me of the old days when we had cell phones, and they would get your cell phone bill, and it'd be about half an inch thick because they would have every call every minute that you made.
Obviously, it's the same thing; you're billing in small increments, which on their own are fairly inexpensive, but once you add them all up.
There is a story where I was working on an acquisition at the time, when we were at IBM and, we were testing something, and my colleague, who was a distinguished engineer, he's now a fellow. We were testing spinning up 20 node Hadoop clusters.
We turn them on, and then we forget to turn them off, and the company we were working with all of a sudden got a bill for a hundred thousand dollars when, and the CFO said, Hey, what is this? This is not in our budget.
And that's actually how we started the initial thinking about cost management; back then, this was about 2015 or so.
In the FinOps analogy, you've got to crawl, walk, then run. Tagging is probably part of that crawling stage; if you don't put it in place, you can't manage what you can't measure. And you can't improve what you can't manage or optimize what you can't manage. You've got to start by being able to measure things accurately.
Which leads me to my next question. I'll give you my perspective when you're finished, but what are some of the common pitfalls or challenges companies face when trying to align their tagging strategy with their FinOps operations?
[00:09:38] Sreeni Maheshwaram: Yeah, excellent question, John, because now everybody realizes tagging is important and then.
If I just go and tag without having the right approach, it is again not going to help much because too much tagging is ineffective. It doesn't help me group and associate my objectives right from a business perspective because, by means of tagging, I'm actually linking my consumption of my asset data to my cloud infrastructure data.
I'm bringing my business context into it. I'm bringing my organizational structure into it. I'm bringing my financial accounting aspects into it so I can correlate all these aspects and make sense of what and where I'm spending, how much I'm spending, and whether it is trending. So, sparse tagging is not going to help either.
Too much tagging is also not going to help. If we consider each tag as a key-value pair, we have a tag and a number of values. Department equals sales, department equals engineering, and department equals accounting; when I put all of them together, it should represent a pie chart as a circle.
And each slice has a certain portion of it, and then I shouldn't be left with a big portion which is unknown. Then it's not right tagging; in general opinion, choosing the right number of tags for an organization is very important without going with a tagging strategy by consulting the stakeholders and identifying their fundamental needs, why we are tagging, what we are tagging for, who's the owner, that's the first and foremost, pitfall we need to be looking for.
The second thing is we need to have the right tagging schema in place with a standard, nomenclature, and taxonomy or the dictionary so we don't invent. Left free for the same thing, people will name it differently: app, application name, application underscore name, with all kinds of mixed cases, lower case, upper case.
So they all mean the same thing, but if you don't have a standard practice taxonomy, even though we wanted to tag, it doesn't solve the purpose. So that's another important thing.
Then the third and most important thing is that it's a continuous process. If we just tag it as one of the activities and leave it, then very quickly, there will be new resources coming in, and very quickly, it will go out of compliance.
So I think these are the three key areas we must keep in mind.
[00:12:11] John Kundtz: That, that's exactly what I've seen in my travels. It's the three Bears and Goldilocks syndrome. You've got somebody that has nothing. They just never bothered, or they went in the opposite direction, as you alluded to, where it's so complicated, and there are too many tags, and the tags don't actually have any business value.
They were put together by the technical or the operations team, but trying to get it from a governance and a cost management perspective it was ineffective. You got to get it just right. It doesn't have to be super complicated, but it's got to be there, and it's got to be consistent.
You brought up the 3rd point, which I really hadn't thought about, which is this is not a 1-and-done thing; you can create a strategy, but you have to continue like anything else because things change so fast today.
It sounds like you've got to continually just make sure it's the right, you've got the right tags, they're consistently deployed, they're actually driving business management results, governance, cost management, things like that.
Not just what data center or server or application does it go to? So that's a great insight.
[00:13:24] Sreeni Maheshwaram: And it can even be measured and tracked. And that's what FinOps suggests in the Crawl - Walk- Run as we mature.
We can shift left and embed some of these policies into the DevOps pipeline itself or use Agile with the providers and have these policies we can attach. So, every now and then, automatically, we can automate them so that it runs and identifies all these violations, and it's the report generated and sent to the appropriate owners. So, every now and then, we know how we're doing on this tagging and the opportunity to fix it.
[00:14:00] John Kundtz: You brought up a good point and an interesting term, which I want to dive into because I'm unsure if all of our listeners are familiar with it. But this concept of shifting left is, I think, really important.
And if you would just so people that don't know what shift left is, if you would give your perspective or your definition of it, I'll give mine, but I'm assuming it's the same.
[00:14:27] Sreeni Maheshwaram: Yeah, absolutely. So basically, in this cost management or the FinOps, we structure in a way. We tag it, and we look at our spending. Periodically, we look at our spending and analyze it and do the allocations, all of that, but over a period of time, once we mature, once we improve the culture in the organization, so everybody follows their own and starts to manage their own, spend by design, then, right from the beginning, like when I am choosing a server, I can already apply my FinOps policies on that,
Am I right-sizing it already? Am I using my budget correctly, And did I tag it well? So, embedding all those aspects at the time of provisioning itself. It's shifting left in this process, not after provisioning, and I started consuming in the operation. So before it even existed, they shifted left and applied these learnings and applied these standards.
So, I avoided the cost in the first place.
[00:15:26] John Kundtz: That's what I'm seeing, if you think about it, in order to have any kind of value in your business, your applications, all this infrastructure that you're trying to manage, whether it be in the public cloud, the private cloud, or traditional data center, it's there to drive some sort of business value, and it's usually some line of business or a business unit.
If you think about it in the traditional sense, you had to have the business unit that was driving the business application. You had enterprise IT that did the management care and feeding of the infrastructure. And then, of course, you had finance, which was trying to make sure the costs were under control and so forth.
As I alluded to earlier in the example, it took one of our clients 45 days to get IT, Finance, and the business units on the same page. And I think as you shift left, it allows the business units who are getting the value or driving it into the enterprise. It gives them more control on both the IT side of the business and the infrastructure side, as well as the financial management side. And they, that way they can react faster.
They have more accountability and, in most cases, are more agile and really deal with what they're trying to do, which is improve their business, sell more stuff, or reduce their costs.
Thank you for diving into that.
So we've hit a lot of different things, and I'd like to wrap this up: for the people who are listening to this podcast and want just to get going, what would be in your mind, the first basic step you'd give somebody considering embarking on a FinOps tagging journey,
[00:17:09] Sreeni Maheshwaram: Yeah, that's a very good question, John. I think I would say the first step is again identifying the right stakeholders and formulating a tagging strategy, which includes a kind of coming up with an approved or agreed set of tags that you know we would want to use. Do a little bit of design around it.
Actually, there is Business-aligned tagging, and there is IT-aligned tagging, so we need to associate what is the purpose of the tags that we identified. And then, who are the owners from that perspective? And, then, from a cost point of view, cost center or a business unit or line of business associating them.
From a classification point of view, maybe identify what kind of data is actually deployed on these servers or the storage components, and then from a functionality point of view, what business services are running on that? What is the criticality of that? Then, of course, from the business naming perspective, what is the service name that is running on it?
So we can classify this in three, six, or seven categories, which basically gives the entire picture of this enterprise, infrastructure, what I'm running, what workload is running, what processes are running, what services are running, what is the significance of that?
I think that kind of gives a good handle. And I have seen some clients even do this very smartly. Because they already have ServiceNow or CMDB established. So, if they have a unique ID for a service, they would use that as one of their tag. So that answers most of the questions. So they have at least a part of it in the tagging.
They can get much more detail going into another system. I've seen people linking it to the SAP IDs and ServiceNow IDs, and some clients who are using the TBM, the technology business management software, which has a lot more organization hierarchies defined their business budgets there.
So, always having that identifier linked into a tag is also something. That avoids defining many tags because you already have something established somewhere you can link to. Otherwise, it's an opportunity to define as many tags as possible, but again, for the purpose.
[00:19:30] John Kundtz: I think you don't have to do this alone.
That's usually what I tell people when you're getting started. It's getting somebody who's been there, done that, helped, and understands what works, what doesn't, especially what doesn't work. So that's another, I think, step that I always recommend to people is you don't have just to sit there and try to solve this problem all by yourself, particularly if you've never done it before, but I want to wrap it up.
If you would summarize to our audience, just why do you think a solid tagging strategy is imperative for a successful FinOps initiative?
[00:20:00] Sreeni Maheshwaram: FinOps is not just about cost savings, the end goal of FinOps is to generate more value out of every dollar that is spent on the cloud or the infrastructure.
So, as a simple use case, having a solid tech strategy solves my cost allocation problem; the industry standard is about 62 percent cost allocation generally, and people are struggling with where other costs are, like who should pay for it.
That's a typical problem statement. It is every client side. So now. Having the right tagging strategy will actually increase it to more than 90 percent very quickly, and there are still some shared costs that need to be worked out using some strategy, but a good tagging strategy will look at that. That's the simplest form.
In a complex use case like unit economics, there it requires multiple tags and correlations, I want to be able to define my Metric, for example, I'm a streaming company and one of the streaming companies. What they have done is they optimized their costs and infrastructure costs, and they went down to a level where they had to tweak the application feature. So that gave them millions of dollars of cost savings.
However, that's fast forwarding or rewinding, the streaming feature. This kind of slows down, which is okay for many users like me, but maybe some users don't like it because they keep going back and forth. They want the same speed, right? But as a result of that, how many subscriptions are lost?
Versus how much is saved, right? So, this kind of analysis is addressed by a unique economic use case. It requires a lot more tags in a correlated way to be analyzed. So that's a complex way. So anywhere ranging from the simplest use case, a cost allocation problem, to a complex use case, to do this, business analysis and statistics.
So, where a company should actually spend more or less, I think that's where it helps.
[00:22:05] John Kundtz: Excellent. Thank you, Srini. This has been a great interview. I want to wrap it up with one more question. And that's how people can learn more and start their FinOps journey. Where would you recommend they go to?
[00:22:17] Sreeni Maheshwaram: Yeah, absolutely, John. So, I always prefer to go to the FinOps Foundation, FinOps.org. And they are very good, the framework definitions and everything are there. And then, there are articles like adopting FinOps. I think that gives a very good starting point. Also, there is a Cloud FinOps book, the second edition, available recently and published by O'Reilly.
I think these two are good references. But more than this, I think one should, as FinOps mentioned, you don't have to wait long enough to become perfect. You should start crawling with whatever you have and incrementally improve upon it. So, I would suggest picking up an opportunity and addressing it.
[00:23:05] John Kundtz: Excellent. We will put links in the show notes for both FinOps.org and some of the other areas and resources that are available.
So, thanks again, Srini. Appreciate it.
My name is John Kundtz.
Thanks for joining us on the Cloud Collective Podcast.
And please don't forget to connect with Sharini directly on LinkedIn because that's another great source of information.
In conclusion, you can succeed with managing and optimizing your cloud costs with FinOps, bringing financial accountability and predictability to the highly variable expenses of the cloud by creating an operating model between IT, Finance, and business.
Thanks again, Srini. Thanks to all your listeners. Have a great day.
[00:23:53] Sreeni Maheshwaram: Thank you, John. Thank you.