Balancing FinOps and Resource Optimization – Dan Ciruli, D2iQ

This is texturing TV. Hey guys. Thanks for the throw.

We're here. We're danceruli who's vice president of product for D2 IQ. We're talking about Fin Ops, which is kind of all the rage these days because well the economy is uncertain and everybody's under pressure or reduce those costs Dan.

How you doing? I'm doing great today. Thanks for having me.

What exactly do we mean by finops these days and I'm asking the question because you know all this Old Timers with a little bit of gray hair scene to remember capacity planning and you know, there were spreadsheets. We were using the track costs and all kinds of fun stuff. So what makes different I am so glad you asked that question because I think in reality there's nothing that makes it any different what it means is that we we need to pay attention to the money that we're spending and I guess the thing that's changed is that it's become effectively.

It's become a more complicated question how much money is something costing us? And and so that's really that's requiring. Well, maybe something more sophisticated and spreadsheets.

But ultimately, you're right. I was laughing as I was thinking about this this morning making my coffee thinking this has been around since there has been you know offsite Computing basically and even when it was on site Computing, you know, how do you how do you balance? How do you think about the money that you're spending?

Either buying or renting these resources and and how that relates to the to the value you're getting out of it. And so, you know, what's old is new again. It is a more complicated than that endeavor than ever before because of the way we consume technology, especially in the cloud.

Well, let's dive into that aspect a little bit because it does seem that part of the issue is we don't really have a lot of adult supervision when people are provisioning Services. We have developers who are just kind of going out and using tools to automate that whole process and then there's a bill that shows up at the end of the month and it's always different and often comes with a lot of surprises. So um, is this a question of technology and processes that we need to think through here?

What exactly needs to be done? It is certainly both and and the way you describe that kind of evolution to the cloud is, you know, it was very organic and and it was very organic in a way that allowed tremendous agility. You know, I remember being at a company 10 12 years ago traditional big IT company.

Yeah, I guess I won't name them but they purported to sell cloud computing. But when I wanted to run a service internally and I wanted to I want to do install our own software and I requisitioned a server. I was told a VM right not even not even a piece of Hardware.

I just give me a VM somewhere and then we're gonna install and run this software. It'll be great. I was told it's gonna take us a quarter to provision it and we're gonna bill you back a hundred k a quarter after that right?

I it was it was amazing with how slow the processes were. So what did we do literally swiped a credit card and we had a VM and Amazon in you know in minutes and so that the way that changed the way you can think about rolling something new out trying something new deploying something new when you could just go to it can just happen right away. So so naturally what happened was individuals individual departments started going out, you know, different parts of different companies started going to the cloud and yes, they were moving very quickly, but what didn't happen is they didn't develop cost mon.

They didn't develop cost controls as they were doing it. They were trying to get that server stood up or and as and as things got more complex, then it wasn't even just service then it's just Services, right? And that's why the the both the the proper monitoring as well as the proper processes around understanding this I of course those didn't know those didn't grow like we those you have to do you have to cultivate those.

How do we go about changing the way developers think about these things? Because let's be honest them from their perspective Cloud resources are cheap inexpensive ordering on maybe drunken Sailors on shortly when spending how do we kind of instill in them some sort of regard for the cost and visibility into that cost structure? so, of course, you can't improve something if you can't measure it and that's the very first thing is to be able to measure something and understand it and it's really important for I for You see iOS to say, okay.

Look the cloud is helping us. Right? It's it's helping us at least with this and this idea of agility and I want to get back to whether or not it's saving us money.

Let's let's move let's come back to that because I think that's an important question but it's it's helping us with agility. But let's ensure that we are are doing things in a programmatized way so that we understand which you know for security purposes. We understand which capability anyone has, you know, you want to control what what you have access to as a developer and ensure that anything that you are using has the proper tagging mechanisms.

So that on the back end we can understand who's using it what it's being used for and how much that's costing. And and you do as a CIO you do have to make sure you are putting in place those strictures because if you're not measuring then there's no way you're going to be able to improve any naturally Is there some way to automate this with policies? I mean can I actually install something that says, you know, if this goes over this threshold send me an alert or turn it off.

I mean what level of control can I get to? Yeah, you can you can get the very good levels of control and and I think that that the the smart cios are out there doing that. There are all the clouds have tagging mechanisms that are built in they are increasing number of tools that you can buy on top of the clouds services that will allow you to understand and and you know, I moved from a company where we had internal controls, but by cost was rarely an issue at Google to a company that was running a SAS and where every dollar really counted and we had to go, you know kind of go back we had we had gone off maybe like a drunken sailor and and just allowed, you know free rein to to Engineers.

We had to go back and kind of retroactively go through all of our old things and ensure that we had properly put in place tags on everything that existed and then put in policies that ensure that going forward we again had those in so that at the end of the month we could get the dashboards that would really tell us because ultimately Don't want to have to go into spreadsheets. You don't have to do any of this manually you want to dashboard and and the tools are getting very effective. Right?

The tools are getting so that the tools themselves will tell you here are your opportunities for you know, here's it here's instances that are outside right where there's extra CPU capacity here's ones where you're not using all the ram you could reduce they can look for patterns of usage that are will enable you to to dynamically provision and importantly deprivision resources. Do you also think that maybe sustainability is going to become an extension of this conversation? Because while we're tracking power we're tracking carbons.

So does this just become another column in my finance dashboard? I think that that's one of the one of the really the things that really go hand in hand. Is that moving to a more sustainable to a more cost effective environment means and more sustainable environment that cost you that you're paying is essentially for electrons and those electrons essentially take carbon and the the more you drive down that that cost the more you drive down that carbon consumption.

So it's one of the the rare cases where it actually goes hand in hand. You can do the right thing and it actually does Cost You Less and I and you know what, it's fine with me when a company green watches and they say hey we did all this for carbon emissions. Well, I don't care if you did it for carbon emissions or you did it to lower your cost.

The fact is you're you're carving consumption is going down and that's a good thing. So it's one of those things where it doesn't cost you more to do the right thing. It costs you less to do the right.

Equal so when you think about this for a minute one of the reasons that people are supposed to be moving the cloud native applications and kubernetes in general is because it enables you to scale up and scale down and yet we hear about overspending and kubernetes environments just as much as anywhere else these days. So do we miss the memo or what's going on with kubernetes in particular? So this this depends on the implementation and I think there's a few ways that kubernetes can help and does help.

If you do it, right. The first thing is that kubernetes in general does because it's spinning up containers that are more lightweight. It does a better job of Bin packing than then VMS do and I knew a company who who I used to work with who when they moved from a VM based environment to a kunettes-based environment had literally a 40% reduction in footprint with the same exact workloads.

Don't serving the same number of customers, but they were able to just using what's happening in kubernetes. We're able to you know, run the same essentially all the same thing and this is an internet Scale Company one of the top 10 sites in the UK, but we're able to by removing the VM to kubernetes and 40% reduction is enormous, right it is Huge reduction and that was without optimizing in kubernetes. It was without saying out we're going to do Dynamic scaling we're gonna do you know, it was just moving to kubernetes.

So kubernetes itself, especially at scale much more and the cloud native ecosystem. I shouldn't just say kubernetes much more efficient. However, we talked before about how cloud-native I kind of grew like a weed within Enterprises in in the devops movement of the of the teens of 2010s.

Everybody was told go do your own thing right start your own cluster everybody go just go do it and and that's really the way devops evolved and the culture evolved for the first five years of this movement. And now I talk to and I have talked to in the last couple weeks. I've talked to cios whose teams are running hundreds of kubernetes clusters.

When you're running hundreds of kubernetes clusters, there's a bunch wrong with that. They have very little control over what's going on. They don't understand their security posture.

There's all kinds of problems. They are duplicating effort right there. People are doing the same thing.

They've got multiple teams doing the same thing over and over and over. However from a cost perspective. You're also not going to gain any of the economies of scale, right?

You've got a bunch of teams that are running small clusters. They're each running a control plane. If they're smart, they're running three and three nodes in that control plane.

So instead of having a few control plane notes and you can get 200 clusters out there. You've got 600 control plane nodes, right? That's not very efficient and you're not getting that economy at scale.

You know, you can't scale down from from a bunch of clusters a bunch of servers to one. If you only start with three worker nodes, you probably need those running all the time. So one of the problems as as teams moved to the cloud where we thought everything was going to be cheaper is that they moved In this kind of cluster by cluster by cluster by cluster way.

I had this cluster sprawl and one of them many difficulties that they're running into is that you now aren't getting those the economies of scale. If you're running in more centralized clusters, then you do have the ability to say, okay now we're really gonna be able to do bin packing more efficiently with with lots and lots of teams and when we do see observances of hey, we really do have a diurnal pattern and you know from from you know, three pm to 9pm is when we're really busy you can actually scale up and you can do that officially and then scale down in a way that actually makes sense and of course can save you money better for the environment Etc. But you can't really do that.

If everybody's running around cluster, right, you know, they're just they're not going to be able to scale down you have too many small clusters running and none of them can actually scale down so you don't gain that visibility, but Going back to you need to understand what you're consuming when you start running when everyone's running their own cluster one nice thing is at least you know, how much everything costs when everything gets into the cluster. Now you need another level of reporting that's going to tell you and that gets a little bit complicated too. Do you think part of the problem is that we have brought a legacy mindset to this environment and people are saying well, I don't want to share it's my own clusters.

So everybody spins up their own cluster and then they just over provision everything because well we were taught to do that because I don't want my application to be disrupted in any way. So give me all the memory you can possibly throw at this thing. I think there's no doubt that that bringing that that Legacy that that Legacy attitude to the cloud cost a lot of money.

And first the first thing we saw was just lift and s*** right where everyone said I'm just gonna take my VMS that we're running in a Data Center and I'm going to start buying VMS by the hour and and what they quickly realize was wait a minute, you know after I shouldn't say quickly after a few years realize that doesn't end up being cheaper, right? If you're running a VM and you're running a VM If you're you're taking it all the time, it's essentially like renting versus buying right in the long run that company's making a profit. They can do some things efficiently, but you're just running VMS.

So that was the first thing that didn't know there was there was not nearly the savings that that people thought and then there there is the kind of the theft and it is important to think about You know, the Genesis of projects like kubernetes was how the internet scale companies were running and and to run something the scale of Google you needed to write Borg and and what Borg absolutely did was Havoc very very strong separation between I'm a developer who's running a service and that service is running on a machine on a VM. You didn't do that that really added that layer of abstraction and that layer of abstraction pays off in many different ways. Some of it is scalability we can scale that thing.

Some of it is reliability, but some of it is there's no Feast. I'm around servers right you you separated that that developer kind of in abstraction so far from the server that you got rid of that idea of I I provision this I own it that the board going it right the kubernetes cluster owns it and so that that separation and that abstraction really is a useful Paradigm. It makes your engineer.

Developers ultimately more productive but one of the many things that it does is it keeps you from thinking? Well, this is my server when I hug the server and put my software on it. No one else is gonna touch it.

Right. So there's all kinds of reasons why that abstraction is useful and we should work to preserve it. All right, folks.

Well, you heard it here chances are if your Cloud bill is spinning out of control. It has something to do with the fact that well they make it easy for you to spend money and maybe you want to think exactly on that process is managed Dan. Thanks for being on the show.

Thanks so much for having me Mike. All right back to you guys in the studio.