Episode 94

Ternary and SADA help solve the biggest cloud adoption challenges: FinOps and DevOps

  • Miles Ward

    CTO, SADA

  • Sasha Kipervarg

    Co-Founder & CEO, Ternary Inc.

  • Lukas Karlsson

    Cloud Architect / Developer Advocate, Broad Institute

As cloud infrastructures become larger and more complex, challenges emerge in the areas of FinOps and DevOps. In this episode of Cloud N Clear, host Miles Ward, CTO, SADA, discusses these challenges with Sasha Kipervarg, Co-Founder & CEO of Ternary Inc., and Lukas Karlsson, Cloud Architect / Developer Advocate, Broad Institute. Ternary, a partner of SADA, is looking to solve these issues. Ternary is the world’s first native FinOps cloud cost optimization tool, built on Google Cloud, that provides organizations transparency to make better decisions and create a culture of accountability, collaboration, and trust between finance, operations, and technical teams.

Host: Miles Ward
Guests: Sasha Kipervarg and Lukas Karlsson

Connect on Twitter:

Connect on LinkedIn:

Transcript

Miles Ward [00:00:00] Hi, everybody. I’m Miles Ward and this is Cloud N Clear. I’m CTO at SADA. And I am super, super excited to have two of my friends who are, right? out at the cutting edge of figuring out the dollars and sense of public cloud. Since, as in like your sense, like are you being even remotely sensible about the way you approach this stuff? Sasha and Lukas are really incredible. Folks, I want you to take some time to listen to their story and how they’ve come to the platform and think through together with us how to how to make sense of your cloud stack. So first, Sasha, maybe introduce yourself to the crowd. 

 

Sasha Kipervarg [00:01:25] Sure. So, Sasha, I’m a CEO and co-founder of Ternary. I’ve got two other co-founders, Patrick and Josh, our chief product officer and CTO. And we started the company last year during COVID. We we left out we’re working together at a company and we left our secure jobs after covid and decided to solve for the pain that we were feeling ourselves at the company we’re at. We had done immigration and run into the cloud budget mess. And, you know, for reasons we’ll probably get into later, we decided to pursue it and it’s been going great. We met Lukas along the way and it’s been a lot of fun. 

 

Miles Ward [00:02:05] I have all the stuff I did the. Of not solving for our own pain and as a result, you’ve never heard of the names of any of those because they have summarily failed. So good good job picking a place where you’ve got a clear idea of what the customer needs as a potential customer. Lukas, maybe the same with you. 

 

Lukas Karlsson [00:02:24] Sure. I am currently a developer advocate and a cloud architect at the Broad Institute where we do genomic research, among other things. And we’ve been using GCP since the beginning. We were one of the first 50 trusted testers of GCE when it came out, I think in 2012 or something like that. So we were early adopters of the platform from the very beginning. And I think part of my story here is that, you know, sort of, unbeknownst to me, a huge challenge. And using the cloud turns out to be dealing with the financial side of things. And so I had no particular interest or skill set in this space previously, but in growing our GCP environment at my workplace from nothing to what is a pretty big environment at this point. Along the way, most of the challenges had to do with finance and communications and things like that rather than technology. And so over that time, I’ve grown to be sort of well known in the space and respect how to deal with building at a massive scale. And at some point when Sasha was looking for some additional people to talk to, Google suggested that he reach out to me. And so I like to chat about this stuff. I wanted to hear what they were doing. And we met a year or two ago. And sort of during that meeting, Sasha, describe some of the problems they were trying to solve. And I pointed out some of the tools they could use to do that. And then I asked a question about like, is this going to be a one time thing that they’re going to do for this one company, or is this stuff going to live on and some kind of product? And sure enough, you know, I hear back from them a while later and it turns out it is going to lead to a product. And that’s why we have this conversation around Ternary. 

 

Sasha Kipervarg [00:04:23] Yeah, and Lukas, it’s a beautiful point to bring up because Lukas is part of our genesis story. Like, we didn’t even realize that it was possible to build a startup in a product around until he asked us, oh, are you guys asking these questions in the context of building something yourselves? And at the moment we’re just trying to solve for our company’s own issues, right? and not thinking of a product. But when he said it, we realized, oh, no one’s built it. And there’s an opportunity here. And I remember distinctly on the plane flying back, talking to Patrick about like, hey, you know, after we’re done solving the problem for a company, maybe, maybe we should do something with this, you know? 

 

Miles Ward [00:05:10] So, I mean, I remember I started GCP at, you know, in 2014 and, you know, I kind of been asking around like, you know, who are the killer customers? And it was like, you know, near arctic in ways. And you were Lukas were the sort of two OGX customers already. And like I talked to Gulas now and they’re like all Myles’s the like, you know, literally I was on a clubhouse last week with Kelsey like, you’re the OG cloud guy. I’m like, no, no, no. You got to talk to Lukas, man. He’s the oldest of the OGs. So so there is you know, there’s there’s a bunch of, you know, benefit from that much history in the platform and recognizing, like, how different it is from, you know, that seven years ago. I like it’s like light years in comparison. You’re in the GCE. There’s a bunch of our customers that have been in each of these sort of different downstream alphas in the, in the, in the run up to, to the sort of conversion over into productization, right? that that transition. That’s super fascinating. There’s, there’s a bunch of spots where, you know, I think the problems that you are working on solving internally are stuff that, you know, even internally at Google, we hypothesized about doing all the time. Like I ended up building the pricing calculator because I was pissed about it, made a bunch of noise, trying to get more help on this set of problems. So, like, congratulations, you absolutely beat me to it. And I couldn’t sucker Google into doing it. But, but I think you’re doing a great job of getting those building blocks done. Talk to me through like, you know, where you feel like you are now in comparison to having just started, like. But what’s what feels good about. 

 

Sasha Kipervarg [00:06:43] Well, I mean, the first thing I should note in and in relation to your, you know, point about Google, like we love the platform and we have chosen it for our migration over AWS for very good reasons, like we felt that not just the technical teams, but the sales teams as well that we were dealing with, they felt very, I don’t know, startup-like to me. They felt genuine and we were getting attention in a way that perhaps we weren’t for made up of us. And we made a really hard turn towards GCP at the tail end of all of our negotiations. And it turned out all to to be good and accurate, like our our sense about what the platform would be. And our development teams loved it. And the only thing that we thought was missing was like how we saw the cloud budget pain. 

 

Miles Ward [00:07:36] Totally. 

 

Sasha Kipervarg [00:07:37] And we went through, like I call it, the stages of grief a company goes through when they can count [inaudible]. 

 

Sasha Kipervarg [00:07:44] You know, like first you ignore it, and then when you acknowledge it, you begin to look around at the native tooling that the clouds offer, right? their own. And then you discover what might be missing there. And then you try and you say, well, they’ve got APIs, I’ll build it myself and then discover, oh, this needs to be your core competency, right?. And it’s a full-time job right? And it’s not like set it and forget it, you know, walk away. It needs to be an ongoing thing. And then, of course, we discover the incumbent marketplace as well. And that’s kind of how we arrived to building our own thing. But like in terms of where we were before and where we are today, you know, in September, we went full time and we were a team of three and today we’re a team of ten. And we’ve got some great advisors as well. We released our product to market in the middle of February, our MVP already, if you look at the feature set, it’s better than all the things on the market. Like, of course, I have a viewpoint, but like the features that we release for GCP are quite unique and we have a weekly release cycle, one week sprints and we release new stuff every week very consistently. When we started in September, it was just an idea. And, you know, today, thankfully, like and we’re very happy that SADA is our first partner. And, you know, we had a very, very short list of companies that we wanted to work with. And SADA was on the top of that list for different reasons. Like some of them you might figure out, some of them might be useful for you to hear, like we wanted our product to operate in a way that was totally transparent to our customers. Like we had dealt with a few companies that come our way at the last company I worked at. And we were shopping around for cloud optimization tools and they weren’t necessarily transparent with their customers. They would sort of hide the margin and you wouldn’t know the customer at, the end customer wouldn’t know what was being delivered quite, um, and so we wanted to deal with partners in a way that felt good to us as well, and we could tell right away that SADA was perfectly aligned with its customers, right? and every person that we met along the way, you know, from John to Heather to Vince to yourself, like all reflected that sort of transparent feeling and that sort of startup vibe, I guess. 

 

Miles Ward [00:10:15] Yeah. 

 

Sasha Kipervarg [00:10:17] And that’s been wonderful. And of course, the fact that, well, you have such a huge GCP customer base and you are sort of on the bleeding edge of of working out all those problems was super appealing to us. So September, there was nothing, no product. And today we’ve got a product, we’ve got fantastic partners. We’ve got design partners at scale as well, giving us great feedback about the features we should build. And of course, we’ve arrived at this partner first approach to marketing as well, like we’re listed in the GCP marketplace and we’re sort of sprinting towards fleshing out the feature set, let’s call it. What we’re most excited about is taking the additional native GCP services like data product, data flow, all the things our customers care about, and first creating insight into those services and then following up with optimizations for the same purpose. So it feels like we’ve traveled a couple of years within a short amount of time, but we’re really happy about it. And I will say it’s an adventure that I should have gone on earlier in my life. But for whatever reason, I didn’t. 

 

Miles Ward [00:11:34] Knowing what I know about the state of the billing API you’d have had to integrate with if you did this a couple [inaudible]. I’m so happy that you did this the year you did, because whoo-wee! 

 

Lukas Karlsson [00:11:50] Yeah, I mean, I think it’s clear that if we had tried to do this earlier, it wouldn’t have been viable because there weren’t sufficient, you know, capabilities exposed in GCP to allow us to build, you know, proper feature set on top of that. So I think there have been a lot of advances in the last couple of years that have made the GCP console very viable for basic functionality. I mean, you know, look, looking back, two or three years ago, I was using Tableau and Data Studio to analyze the data. I mean, there was a time before the Big Query export was available and I was trying to do all that with CSV files that were coming once a day. And so the evolution has been huge. Now we have not only the, you know, the ability to look at the the GCP data directly from the big query feed that’s produced by Google. But there’s now these budgets and a budget API that you can integrate with and the budgets now have the capability of reporting to pub subs. So that’s another way to integrate with the platform. And Google has exposed a suite of recommendation APIs that sort of give us insights that we can then use to produce even better insights for the customer. And I think another thing about Google’s approach is that so much focus around the machine learning aspect of it, like instead of just being like, here’s your bill, you know, they’re like, here’s some instances that you’re using that are bigger than you need and you can save money by reducing the size, for example. And so if we can take those at scale and try to produce useful interpretations of them and communicate back to the user, then that’s really powerful and it makes it very seamless for the user. 

 

Sasha Kipervarg [00:13:48] Yeah. And, you know, it’s not just the right timing on the technical side. It’s I’d like to give like a shout out to actually the incumbent competitors in the space. Like, this is kind of a brand new market. And companies like Cloud Health and Cloudability, prior to acquisition, like they pioneered it, right? like folks like JR and Matt, and I think we owe them a debt of gratitude because they were doing this before anyone else, before there was a market for it, even, right? And then most recently, I think what I’m most thankful for in terms of timing is the FinOps Foundation. You know, it moved over to the Linux Foundation. It’s an independent entity and they are creating best practices and offering them up by API and have a lot of plans around that. And they’re making it, I think what’s close to an industry standard, and I think were it not for all these things happening, you know, all the work that Google did, all the work that that you both did in the past as well, Cloud Health, Cloudability, the FinOps foundation. I think we’re sort of standing on the shoulders of other people, right? The work that they’ve done. So I very much appreciate that. And I’m very aware of it, too. 

 

Miles Ward [00:15:01] Yeah, 100%. I mean, I spent I was early early solutions architecture at AWS. We get this security notification that there’s a whole bunch of root credentials being used out of Portland, like across multiple different accounts, for like oh man, somebody has hacked their way in and is stealing accounts. I was also at the same time like building up this copy of a machine to mine Bitcoin, because that’s what people were doing when they hacked into Windows instances. It was like all of these like dark days at the beginning where you’re trying to figure out how to keep people from stealing cloud. And so I literally drove to Portland and had Matt and JR and those guys beat up this whiteboard about how they were securing credentials, something like if you had to do all of the hot violence they had to do to be able to even get access to this data, it’s just a huge burden to bear as a startup or like like all of this just cruft that, you know, clearly nobody wants to have to fight with. So I’m very, very glad that you get to just get into the hot functions that the customers actually need. And I think that the need there is bottomless. So we represent this, you know, a huge percentage of relatively of Google Cloud customers. And, you know, I get asked all the time by Google product managers like, hey, do they want more features and Big Query or are they, like, really geeked about, like, performance specs and GCE or, you know, what feel they think they have for criminalities? I’m like, they don’t care about any of that. They want the bolt and nuts. Could you make it so that they understand what the stuff costs, right? and and I think that there’s… It’s one of those places where, like, if we could describe to Googlers, that the bill being solved requires them working around some of the basic laws of physics or that it’s like a daunting, multigenerational computer science problem. You know, they would attack it and there’d be like this rabid assault on on the sort of problems facing they’d be all over in a week, right? Like spanner breaks all sorts of basic preconceptions I have about how data works and they just try to bang that thing out in a couple of years. Well, the bill isn’t that way. It’s just hard work. I you have to pay attention to the nuance and the detail. And I think our customers expect that as a minimum barrier to entry. I’ve talked to a lot of companies, you know, about this, the big numbers, right. There’s like three and a half trillion in total infrastructure everywhere, and maybe less than 10 percent of that, about 300 billion is in public cloud now. And if everybody sort of generally agrees that cloud is better and it seems like there’s, you know, more than 10 percent of the workloads that made it to public cloud, there’s some delta there and that’s the aggregate cost savings of cloud. The impediment is figuring out how to shove it into the 10Q filings of every public company. And I mean, it makes sense in the budget process. So, right? maybe go down one layer further for folks. I mean, it’s one thing to say, like we help you understand the bill. Like, is there any extra detail you can give about, you know, what kinds of problems they might be running into that are product like Ternary helps solve? 

 

Lukas Karlsson [00:17:54] I can tell a little story, actually, if you like. So, you know, where I work, we have a large GCP environment and we use other clouds as well. So it’s not just a problem. 

 

Miles Ward [00:18:07] He’s not joking…extra large. Wasn’t really like trying to debate, like how much more quickly you could delete petabytes of data from GCS. 

 

Lukas Karlsson [00:18:17] And we like deleting data. That’s a that’s a fun thing to do. So but we’re on the order of, you know, tens of thousands of GCP projects, hundreds of thousands of buckets. And so there’s a lot of I guess the further detail I should mention is there’s a lot of funding sources. We’re not just paying our bill with a credit card and that’s, you know, the company budget for it. It’s being paid out of hundreds of different grants. So it makes the billing problem sort of more challenging because it’s not just one billing account. It’s whatever your problems you have on a billing account. We have that hundreds of times over. And so one thing that’s really challenging is Google now provides these cuts. You can you can purchase these CUDs (Commited Use Discounts) and there’s a API that provides you with recommendations on how you can consume these. And so for me, managing a large environment, you know, I have thousands of projects, some of which can benefit more from CUDs and some not so much, and sort of making those decisions around where should I focus my attention to get the best bang for the buck. That’s pretty challenging at the scale and requires me to write a lot of code and to kind of piece through the data. But in Ternary there’s a feature where I can go in and sort of drill down on a region and an instance type, on a project and see the the project that is consuming the most compute resources. And then they provide a visual tool that allows me to sort of move a bar, you know, one side to the other to sort of make a decision or an experiment on what if I purchased this many cuds for one year? And what if I also added in some CUDs for a three year agreement as well? Like, what would that look like? And so I can run those analyzes right on the screen and come up with a real understanding of what it would look like and then make that decision right from there. And so it’s like they put the time in to analyze that data and to figure out a way to present it to the user that makes sense and is easy. And I don’t have to write all of this code in order to understand what’s going on in my environment. I can simply point someone at this visual and say, like, look how much money we could save if we made this decision. And is there a different kind of decision that you’d prefer to make, like let’s, you know, tweak these dials a little bit. And so we get it to look the way you want and then we’ll see what that looks like. And honestly, I don’t know how I would have accomplished that goal prior to having that tool. 

 

Miles Ward [00:20:56] You’re exactly right that that people are horrific creators. They’re great editors, right? if you say this is what you’re going to be able to see, this is how much we would save if we make a choice, then it’s quantified like, oh, yeah, I want to have that meeting. I want to understand how we straighten that out. Where before if it’s hypothetical, maybe, through some analysis, it goes nowhere, right? so and there are you know, I always get frustrated, you know, like cell phone plans, right? I’m like, why do you give me all these options during signup time? Just like, watch my usage and charge me the right amount. You pricks, right? like, there’s just no reason that it should work that way. But there are so many different pricing models and if there’s anything that I’ve seen, it’s not like the number of models or the level of complexity or the number of SKUs in total or any of those things are going down, going up, right? like they’re going to give you more and more options. I think it’s the right thing for the cloud providers to do to just ship primitives like wild animals in every different packaging and sorting that they can figure out how to make money with. And it’s a perfect counterpoint to them to have a product like Ternary in place to help you think through the optimization dynamics and make sure that you’re capturing the opportunities that exist. 

 

Lukas Karlsson [00:22:08] I mean, even if you look at GCP, when it first came out, you had these extended use discounts, which was super, super easy for the user to consume because it was just like do what you’re doing and we’ll give you some of the money back afterwards if you do enough of that. So the sustained use, exactly. And so it’s like, oh, I use the machine all month and then I got a credit because I did that. Thank you. And so there’s no effort. You just get free savings, which is nice. But then there’s some folks that are like, but I know I’m going to use these machines for a really long time. I want an even bigger discount. And so then they add this feature and now you’re faced with this decision to make on what’s the best way to approach your particular problem. So then you start to need these third-party tools to help you make a decision between all the choices that the vendor is providing. 

 

Miles Ward [00:22:52] Yeah, and I think it’s. Do you feel like, you know, maybe this is more for Sasha, like there’s on the one side we are, you know, I think the tools that are being built now, you’re a tool in specific is, you know, sets up an additional layer of viable buyers, right? like makes it so that another class of customers can successfully navigate the morass of public cloud and come out the other side of something that’s successful and works with inside their business. But the next round pass that is even bigger, right? where you get to tighter controls, not just analysis around this stuff. I think one is a necessary precursor of the latter, like not going to be able to automate this stuff if you don’t have the data to begin with. What you know, how do you see that sort of maturing over time? What you know, what do you think companies and businesses need to do to to kind of get ready for that change? 

 

Sasha Kipervarg [00:23:52] So at a high level, I think what we’re talking about is eventually getting at the cost of a company’s widget, right? like not only for their cloud spend, which might be their largest spend, but like how much that would have cost across all their other systems, right?. And I think our product is a step in that direction. And certainly, there are lots of companies pursuing it. But like at a high level, philosophically, like Ternary is about helping companies make the most effective trade off decisions. Lukas was just talking about, like having discussions and our essential insight that came from our own experience and our own pain and from talking to other members of FinOps Foundation who are having cloud budget problems as well, was that the missing element in the marketplace today was actually collaboration. It was, you know, the algorithm  surfaces a recommendation, but ultimately to complete it, to get engineers to take action, you actually have to have a conversation between finance… 

 

Miles Ward [00:24:49] You can have the finance people want  to do it all day long, but they’ve got to go to and execute  impossibly technical thing over here. And if they click wrong button, they’re screwed, like now you’ve got to figure out how to submit a ticket. They never submitted the ticket before they back something out, like the operational implementation of the asks that they have, that are well-founded asks, is so foreign structurally to the way that… 

 

Sasha Kipervarg [00:25:11] Yes, yeah. 

 

Miles Ward [00:25:12] ..send them a PO. NO, you do not send them a PO, you click a button on the website, right? you know, like, I don’t want to do that. 

 

Sasha Kipervarg [00:25:18] And let’s zoom out for a moment and talk about what’s actually happened in the background, because that’ll be useful context for the rest of the conversation, right? What’s happened is we’ve gone from an environment where on-premise we had capital one-time expenditures, right? and very little operational cost. And it was the high priest and finance and engineering who would say, let’s buy a whole bunch of racks of servers and implement them. And then once we’re done, it doesn’t matter how often the developers use them, as long as they don’t top it out, it doesn’t cost this extra money. And then what we said was, hey, DevOps is great, SREs are great. We want engineers to be able to scale the product really fast and have an API to spin up infrastructure, which we all agree with, which we know makes a company go faster. But we didn’t give them the tools and the knowledge to know that what they actually had was an API to spend money. And they were the decision-makers all of a sudden, right? And so, of course, we gave them a Porche and, you know, they ran it off the rails. 

 

Miles Ward [00:26:19] They floored it, floored it! 

 

Sasha Kipervarg [00:26:22] You want to go fast? I’ll give you 10000 CPU’s right now. Right. But of course… 

 

Lukas Karlsson [00:26:27] It’s like an API that spin-up credit cards, basically. 

 

Sasha Kipervarg [00:26:30] Right, exactly. So this is part of the shift that happened. And then at the same time, because the shift happened, finance needed visibility on an ongoing basis and they needed to be in the conversation to purchase things like CUDs, long term contracts, right? You go to an engineer, you know, they’re not in that mind frame and they don’t know about the budget necessarily. It’s finance that has that context. Really, it’s like cats and dogs needing to work together for a company to be successful in the cloud right now. And let me just throw some statistics at you that maybe, you know, maybe you don’t. But like 80% of companies that are operating in the cloud overspend their budgets. And when they do, it’s like 20 to 50%, right? And you mentioned before, like where the infrastructure market is today. What the statistics I’ve heard is like it’s a roughly $120 billion-year market today, projected to be $500 billion in two years from now and then trillions a while, because maybe only 5 to 10% of infrastructure spend is in the cloud yet, right. But it’s all moving in that direction. 

 

Miles Ward [00:27:38] Yeah, I think the three hundred also grabs like all of the platform as a service, workspace and needs and all that kinds stuff. 

 

Sasha Kipervarg [00:27:47] Makes sense, makes sense. And so you have a train going down the tracks. It’s going to go like off the cliff unless you put in the guardrails, right? And I think it is the case that companies have to operate in a different way. They need a different operational model for the cloud. And I think this is the idea behind FinOps, right? It’s this recognition that you need this collaboration between finance and engineering. You need the best practices and you need ongoing insight and optimization to be successful with your budget. So that was our understanding at the last company I worked at, it’s the pain that we had, it’s the pain that we’ve heard from all the other folks who managed cloud budgets. And when we built our product, we built it on top of those principles, right? that’s how we arrived at having collaboration features at the very heart of our product, right? So when we surface in optimization, what we do is we allow you to assign it. We allow you to have a discussion about it. We allow engineering to push back on it. We allow you to say, hey, this doesn’t make sense. Here’s why. Here’s the context that gets tracked with the optimization. Here’s the synch into Jira ServiceNow. And we’re just going to, we’re going to defer this for six months because it makes sense to do, you know, after six months, because the contract required us to keep this multi-petabyte file, you know, until X date, right? So it leads to really effective conversations. Where I think we’re going is once we have a really good handle on the cloud and the budgets, and once we have shifted a company’s culture so that engineering and finance are working more closely together, our engineering, finance and procurement, perhaps. Then I think we can begin to bolt on additional data sources and we can begin to sort of link all these numbers together, right? like imagine that you have a, I don’t know, you’re spending X, Y and Z on some other products that are not part of the cloud, but they make up the cost of your widget, right? And you begin to pull that information in an aggregated in one tool, a source of truth for these conversations, and you begin to assemble what the cost of your widget is. I think that’s the Holy Grail that I think most companies want. 

 

Miles Ward [00:30:05] Yeah, there’s a bunch of this that it’s difficult because I think companies you describe to them the effect of the controls that they seek, right? like, I want to have an explicit this thing’s never going to cost more than a thousand bucks, OK? When it goes over the thousand, I shut it off and data and they go, oh, no, clearly, no, that is not OK. Right. And I go, OK, so at a thousand bucks I put your data into like a public repository that lets other people read it and write from it for like research purposes. Like, no, clearly no, that is not OK. Right. And so, like understanding what the real bounding box and what actual operational characteristics they want their systems to have under cost pressure in the same way as it behaves. Or they have to understand the behavior under performance pressure or traffic pressure or scaling pressure or security pressure, right? it’s just another dimension in those in that optimization function we’re all trying to work out. 

 

Sasha Kipervarg [00:31:08] Yeah. 

 

Miles Ward [00:31:08] And I think that there’s a lot where when we can start to talk to finance about this from like a risk mitigation and risk management standpoint, I think there’s a lot where now we’re you know, we’re really talking their love language and we can really get into the work. Because those are dimensions of control that bare metal and sort of physical infrastructure just do not have, right? And it’s it’s one of those spots where I think the better we can enable them to identify and enumerate and mitigate risk. 

 

Sasha Kipervarg [00:31:41]  I’ve got, I’ve got two related points I wanted to bring up. First was, you know, like when the DevOps movement first came to be, right? and like it emerged out of agile, right? This idea of a full-stack developer was extended into someone who also understood infrastructure and someone who could deploy and support, get feedback on to fix the product, right? So we sort of extended our thinking about what a full stack developer was. I think we need to extend it just a little bit more now. Like, we need to make engineers care about their application’s cost efficiency, as well as the dimension of performance, right? because until they start doing that, until they end, until someone comes to them and gives them the insight they need to know what the cost efficiency actually is and then gives them recommendations, it says here’s how you can improve it and then maybe benchmarks their applications cost efficiency against others. We haven’t given them the tools that they need and we haven’t let them know that it’s important. But I think in order for companies to be successful in the cloud, I think they need that the engineering side of the house needs that on the other side of the house, the finance folks, they need tools to understand the data in a way that they can best absorb it to, right? they need the same cost information expressed through budgets and forecasts and actuals and CUDs analysis and all the stuff. That allows the two sides to come together in the middle to have really effective conversations, right? And this is sort of like what our tool is doing. It’s what I hope competitors build into their tooling as well to really help the industry. And I think that the opportunity for companies who use the cloud is that they can build their products faster. And if they build them in such a way to take advantage of the scaling mechanisms of native services in the cloud, then they can arrive at a lower cost for their clouds, right? But they really have to be thoughtful in their approach. They need the tools and they need that culture shift. 

 

Lukas Karlsson [00:33:46] I think one thing that’s really changed is the number of different people or roles that can impact the cost now, whereas before, you know, with the with the physical model of buying assets, you would have like a finance person who gets the final sign off on whether a purchase gets made or not. And everyone else is kind of invisible to that until the decision is made and the hardware arrives and the engineers are like, cool, we can use that now. And, you know, there’s really no issues in that flow. But in the new model, you know, you have the finance people. And Miles, to your earlier comment, I would say if you asked one of our finance people, so when it hits the ten thousand dollars, you just want everything to stop working right. At that moment, they would say, yeah, like whatever is going to make the money, stop spending. That’s what we want. And then you ask the engineers who built the infrastructure or the PI whose research is being done on that infrastructure. And they’re like, no, no, no, I don’t want that. So different people do want different things. The finance people are fine to have a hard cap where things just stop working and the engineers don’t have a sufficient amount of insight into what’s going on to prevent it from hitting that cap. So that’s where all these tools and communication are needed. And to Sasha’s point, I mean, as an engineer who’s building infrastructure, I need a certain amount of insight into the billing in a way I never needed in the previous model. I never had to know who was paying for it or what the budget was. It’s there, so I use it. And now, like, if you go into the cloud at the same model of like, oh, I can create it so I can use it, there’s a huge disconnect there. And that’s the problem I think we need to solve. 

 

Miles Ward [00:35:33] Well, there’s we keep seeing like there’s a product on the Google side called Contact Center AI, and one of the things that it has is agent assist, where it would be one thing if you had, like a robot talking to people on the phone, but if you have a human talking to people on the phone, like help them out, like tear after them and give them suggestions and best practices and guidance. And when you hear stuff from the user, there are questions that prompt response, you know, go tear this stuff down. It was like immediately in that demo I’m going like, where is the agent assist for the building console?, right? like, tell me what stuff sucks and where I should turn it off and how to work this out. Like, you know, I don’t know if they want to like name the agent assist, you know, query or, you know, Corey Quinn the Younger or you know, they would make fun of somebody about it. Right. But, but generally like having, you know, having something even just right along sidesaddle and help out. I know there’s a bunch of work going on on the Google side. They they they sort of rotate through different labels. Sometimes they call it self self-driving cloud. Sometimes it’s autopilot. There are a couple of others… 

 

Lukas Karlsson [00:36:35] Active Assist. 

 

Miles Ward [00:36:36] …that they talked about. Active Assist is another one that they used. I will see when they like – foo-woo – net down to a single label. 

 

Lukas Karlsson [00:36:41] To think that an Active Assist came out of a meeting I had with Google. 

 

Miles Ward [00:36:46] Aaah, there you go. 

 

Lukas Karlsson [00:36:47] They came back to me were like, yeah, we picked a name. We wanted to know what you thought it came out of this thing you said one time. 

 

Sasha Kipervarg [00:36:51] Is this like the ontrack system for Google Cloud? 

 

Lukas Karlsson [00:36:55] Yeah, the self-driving cloud. 

 

Miles Ward [00:36:58] I mean, like the the predecessor to that, the project at Amazon was called Kumo and the output of that is Trusted Advisor. I proposed that. So like I stood up and said, you know, this is stupid, we know what customers ought to be doing, we should just sort of tell them in situ. So I was proud to hear that Google was doing sort of run after that. And frankly, like rather than to derive that stuff statistically or by way of a rules engine, to use this handy dandy machine learning stuff that we’ve heard so much about to try and get to clearer and more accurate suggestions, which from the examples that I have seen so far, worked really well. But it’s all of that that I’ve seen so far is designed around this very narrow view of the user, right?. So when you’re talking about like there are now lots of users, boys and girls, like we have to handle the people in finance and procurement and legal and compliance, and they don’t they have a totally different user profile than our friends over here in development. And and, you know, the assistant suggestions of that person are not going to make sense to this person in our current actions that they can follow up on. So there’s there’s a bunch of this that’s, you know, ripe for optimization and, unripe for better communication and better transparency and education. All that, you know, makes for smart folks like you working on the problem very, very happy for me, it’s a wonderful thing. 

 

Sasha Kipervarg [00:38:14] And, Miles, I’m curious, like from your view of it, having, you know, you worked the US worked at Google, now you’re at SADA. What is your view about what the natural role for the cloud provider should be in this sort of market and what the natural role for like a third-party tool like Ternary should be like? How do we all naturally align together to solve for a customer pain? 

 

Miles Ward [00:38:38] I mean, I think there’s…you know…, I think there’s a model of communication about the telecommunications companies where you would say that they are just dumb pipes and then, you know, the service offerings that they put on top, you know, represent this sort of collusion effectively where they are optimizing themselves. And that’s silly. And, you know, you don’t trust them to cut your bill down. They are the ones whose bill you’re heading down. They’re never going to be as aggressive about that as an independent third-party. I think there’s a lot to having a check and a counter and someone to hold accountable the cloud providers is one of the things we get asked by customers all the time. Are you sure they built us the right amount? You know, right? I’m like, well, you know, you consume several hundred million iOPS of this kind of storage. You know, I’m not sure to how to tell whether you, like, accidentally did 3%t too many of those or or, you know, those were underreported by the. That’s really, really now you’re talking about in many cases, especially for systems that are sort of small and optimized, the cost of measurement of that might exceed the cost of the system itself, right?. So you’re in this weird arms race with a very giant competitor in the central cloud business. 

 

Lukas Karlsson [00:39:52] And there can be mistakes, like I mean, I had a situation not too long ago where, you know, we got a strange billing situation. We looked into it. It was a new service we were using. And when they provisioned the service in the platform, they mistyped the price code off by one. And so we were getting paid the charge 10X what that service was supposed to cost because they just like mistyped it in the pricing API. 

 

Miles Ward [00:40:21] Those little zeros are pesky. 

 

Lukas Karlsson [00:40:24] Yes, you do want to have some way to like have, you know, a third-party view or a different view of the of the bill that allows you to compare it to. I mean, a lot of us are not just in one cloud, right? as well. And so there’s there’s the idea of being able to compare, like when you’re doing something online and on the other like are are these adding up like are these competitive prices that I’m seeing between providers or are they drastically different? And that concerns… 

 

Miles Ward [00:40:53] The providers are insane, right? like the providers have gone apeshit with a zillion different nuances inside the pricing system in slightly different assumptions between them. I don’t know how much time I have spent doing horrible, horrible things in spreadsheets to try and normalize in these systems, right? like just do S3 to GCS, right?, I mean, you’re talking like minute differences in the behavior of those APIs under different conditions. And the pricing units are also ever so slightly, is it petabytes or… 

 

Lukas Karlsson [00:41:24] Yeah, like, are you going to use the code line and how frequently are you going to pull it back. Yeah. 

 

Miles Ward [00:41:29] Yeah, totally crazy. How fast will you pull it back when you pull it back. Right. Like uh so stuff like that, there are problems. They are complexity that impedes adoption and they’re naturally best solved once, right? like you almost want to like layout to every future cloud customer and say you should invest in advance. I’m just going to make it so you can actually adopt this by the time you show up, ready to go to use this thing. Right. Because there’s so many companies that are just they see the marketing, they’re excited about adoption. They’re getting into it. They’re like, yeah, as long as you fit into my body, this is going to be great. And like, it won’t ever fit in your budget, bro. It’s very strange. So that’s a that’s a gnarly, gnarly world. You know, the good news to them is it makes them faster and in aggregate cost less. But, you know, alignment to planning is not something everyone like does. 

 

Sasha Kipervarg [00:42:22] My view of the of the relationship, like when I first started figuring out, like the startup in the problem, I kept getting the question from potential investors, like, well, won’t you know, Google or Amazon or some other cloud, you know, won’t they continue building their tool? And then eventually people will need a third-party tool. And also they mentioned the dynamic about, well, you know, your cost overruns are the, you know, cloud providers profit, right?. And it made sense initially. But then when we thought about, we realized, well, number one, the cloud providers know that their customer spend is always going to go up anyways because they use more of the cloud, right? even if they make the best trade off decisions, it’s always going to go up, right?. And so they don’t want to piss off their customer. Right. And so they are actually naturally incented, especially now that there’s more competition between clouds to actually help solve the problem. And then additionally, when you when you think about features like, I don’t know, anomaly detection, right? like imagine someone speeds up 10000 CPU’s because I don’t know the their telephone code was off by one digit or something like that. And all of a sudden you begin to pay for them. Right. Today, most of the tools in the space, they they work off of the 24 hour billing file, right? But if you want real time insight, maybe you want to work off of stackdriver logs or data. But to do that, you need to ingest the company’s application logs, right?, which is something that I think most companies would not want their cloud provider to do. So it feels to me as if for features like that and there are many others like integrations with Jira, ServiceNow, the budget stuff, like there needs to be a third-party Switzerland, for this, right?. And of course, I’d love it to be Ternary. But like, I think it it’s a natural fit for the ecosystem where third-party products can do this on behalf of customers. Cloud providers can provide the API and the basic infrastructure in the basic bill, right?. And then, of course, the natural ecosystem with partners to, right? that I think it all works really well. 

 

Miles Ward [00:44:31] I think it goes from a very sound idea and probably an important division of responsibility and the sort of correct check and balance thing, when you’re managing one cloud, it becomes. Absolutely critical when you’re at the multiple cloud vendor layer, because there is zero percent chance I am going to send my Adewusi bill to GCP to have them do the comparative analysis, if only because I want to fight them against each other for cost purposes, like this is just a hard no, absolutely I’m not going to hand them that data right now. You can ask any employee at any company anywhere and they will tell you that’s probably a bad idea. So it’s you know, it’s spots like that where, you know, Switzerland is the right thing. And I think that there’s a lot we’re thinking through. You know, what we absolutely imagine as the next major phase of all this stuff where an increasing ease of use and simplicity of adoption means you’re going to get not less frequent movement between the problems of a more frequent movement between the platforms over time and more interop and more simplicity. That’s the just sort of general trend and always has been. Hmm. It you know, you’re going to need to manage that or you’re going to be an even more risk than you currently are in your one or two providers that you manage independently. 

 

Sasha Kipervarg [00:45:52] So, Miles, you you and SADA have this treasure trove of information sort of sitting on, right?. And what sort of insights do you currently have about customers moving sort of between clouds or like the desire? Are there any interesting statistics there? 

 

Miles Ward [00:46:09] Yeah. So you are you are presaging one of my OKRs for the year. We do have a healthy pile of data. We, of course, have to do the Google thing and organize it and make it useful and accessible. We have it. So at least like that. And and I will tell you, you know, we are very focused on one of those trajectories, like moving from other things to GCP, because that’s kind of where our bread gets buttered. And and, you know, one of the things that’s been pretty surprising for us, like all of the preexisting expectations about like how long it takes to get started and, you know, what is the sort of average time for developers who are interacting over, you know, their source control to start submitting patches that look like they’re the right kinds of things in a new cloud environment versus the other ones. Like there’s a bunch of spots like that where I think the general terror about the human switching costs is a little overblown. And then the other side of that is like, you know, what we’re watching is like the technical how long it takes just for the bytes to flow around and to get the zeros and ones from this thing over to this thing and to actually delete from that thing over there. That seems to be taking longer than a bunch of the estimates that we’ve had held on. So that there’s some spots where, you know, we’re not only are we surveying the data that we’re getting from customers, we’re surveying all of our engineers and they’re their customers and engineering and thinking through kind of like, what do you think this ought to be? You know, what’s your guess about this number? Do a little wisdom of crowds bid on it? And the comparisons of actuals versus those guesses are that’s one of the spots where there’s some pretty startling discrepancies that hopefully will help people move more quickly. Like I was super inspired. Our whole team has been by the door metrics and how, you know, how clearly, like, if you go faster, you’re going safer too, right? as a as a sort of general guiding principle. That’s I think there’s going to be some spots out of the data that we’ve got and then hopefully we’re working on together that will help people come away with the same kind of like really. OK, I’ll give that a try. 

 

Sasha Kipervarg [00:48:14] Gotcha. And Miles, I’m just curious, like at a high level, what what was what were the signals that made SADA go all in on GCP? I mean, you must have seen extraordinary stuff. What was it? 

 

Miles Ward [00:48:28] Well, so a couple of different pieces. You know, Eric Schmidt has a great quote where he says, you know, revenue cures all ills and growth ain’t bad either is the sort of back half of that sentence. And, you know, if you’re going to be a partner, if you’re going to be a Remora on a shark. Living shark there is and Google is growing faster than the rest of them, and I think that our position as a partner for customers is that I’m I’m taking the bet together with them. I’m on the hook for the operations of their platform. And I have seen more in SRE and operations, reliability and the way that the core systems are operated, where I feel just dramatically more comfortable betting my company’s future on the operational characteristics of GCP like I’ve been in the DCs for Amazon, I’ve been in the DCs for Azure, I’ve been on all these different spots, like one of those three looks like the Starship Enterprise and is built out of shit I’d never seen before. And so the results are, you know, are in are in the pudding, right? the status page speaks for itself. So every one of these systems, everybody goes down, everybody takes downtime. But I also hugely appreciate the degree to which core engineering at Google, you know, is, you know, totally transparent and utterly direct about the postmortems. And the analysis thereof is like half of those are college courses on their own about what scale operations look like and what kinds of problems you can run into when you’re in these kinds of macro scale, interdependent systems. I know a lot of customers that spend a lot of time chewing on those those documents to kind of best understand what to anticipate as they get bigger. And that’s that’s the team I want to be learning lessons from. So it made a lot of sense for us to bet that way. 

 

Sasha Kipervarg [00:50:18] Yeah, I think Google has the benefit of being, you know, third to market because of all that time to basically observe how it was done before, figure out a better way to do it. 

 

Miles Ward [00:50:30] And they’re cheating. They’re utterly cheating. They had a functional high performance, scaled global, private, but effectively public because you got how many zillion divisions inside of Google working cloud functioning, abstraction layer and an API. They built the cloud on top of the cloud are like, if you’re an Amazon, you had to build a cloud from scratch, like with computers, right? like much harder problem, frankly. So, you know, got got to got to give credit to to the fine folks and Colossus and Borg and big table like some of these precursor building blocks are super super crucial. 

 

Sasha Kipervarg [00:51:07] Yeah. Totally. 

 

Miles Ward [00:51:09] Spanner. 

 

Lukas Karlsson [00:51:12] It shouldn’t even be possible. 

 

Miles Ward [00:51:15] You know, I talked to this Spanner product managers last week, and, you know, I’m telling you about the stuff like, yo, you got to make the pricing right and you got to help people understand what the prices are going to look like. And you’ve got to, you know, boil this stuff down. And like all the answers or the right answers, go guys go, like very, very excited about where they’re headed this year. Should be fun. Like looking forward to seeing more of their more of their data in my system. 

 

Sasha Kipervarg [00:51:39] Miles, I’m curious, like GCP is an amazing shop and I agree with everything that you said to, right? like it’s part of our rationale for landing our ship in GCP ourselves. Right. And what what inspired you to to go to SADA? 

 

Miles Ward [00:51:54] Oh, totally. I, I had I had seen this movie before, so I worked at Amazon. I helped build the partner programs, their work together with Tom Stickle, super bright guy. You know, Dorothy was amazing on the resource side and that company got over eight million, a billion in revenue, went to sort of a direct, you know, quarter bearing technician sales model, brought on a bunch of sort of regionalization to make it so that that worked out, had a bunch of capacity outages as a result of sort of unexpected organic growth with customers and as a result of refocusing on the largest customer opportunities. Because the only things that the news would pay attention to were billion dollar deals. You shed a bunch of work down to small partners. Fast forward five years. I met Google and they’ve gotten to eight billion in revenue and there’s a regional sales model and they’re going to based compensation for everybody. And it turns out they’re having capacity outages as a result of organic growth. And, you know, they’ve got this whole thing, like, I think I know how this is going to work. So part of it is just sort of like, you know, looking forward to a bitchin rerun of a great TV show. Another part is I had gotten up to a team of 85 people and I wanted to be hands on with customers now. And there wasn’t really a way for me to shed that many folks. I love them. They’re incredible team members. It was not them. It was me. It’s hard managing that many humans in a big environment like that. And so as a startup or and someone who’s sort of generally humanly more comfortable at this size, this was like a much more much more comfortable place to go. And and Tony and Dana and Narine and Orkideh, the leadership team here are incredible. And they were like a team I wanted to be a part of. So it was it was an easy choice. 

 

Sasha Kipervarg [00:53:46] It’s a cool story. So, you know, related to the partner comment, like a part of our shtick here is that we went partner first because we recognize that partners already have established and trusted relationships with their customers. And we didn’t want to get in the way of that. And, of course, you know, we’re a software company. We know how to build software really well. We suck at professional services. That last mile problem is not something we wanted to work with because it’s not a core competency, right?. And so we thought, like we had heard from one of our early advisors and VCs also, that it was really difficult to bolt on a partner program after after you started selling because you have a particular model, maybe hire a large sales team, yada, yada. And we didn’t want to do that. So right from the start, our strategy was we wanted to go partner first, which meant we would take advantage of their own relationships. They would maintain them themselves. We would just make them great software, not only for the customer, but for the partner as well. Right. So like we have these consoles that we’re building just for partners, we are taking feature requests from partners such as yourself. And I don’t think we ever want to get into the business of, like having that direct sales motion. 

 

Miles Ward [00:55:07] Channel conflict of the thing, like, you know, and I think that there’s a lot of a lot of businesses who kind of misunderstand our layer and our part and, you know, not at their peril, but like to the benefit of the businesses that have been able to successfully work with us in this way. And, you know, I mean, like great customer of ours, public case study, super, you know, big obvious brand name. Everybody we know about, you know, same rep, same technical team on our side for five and a half years. They’re on their seventh Google contact, we’re just like, that’s the way Google needs to operate to do the things that they’re doing. I totally respect that approach. Customers dislike it just generally. And so the more we can do to be the stateful interface, the the trustworthy component in that structure and frankly help them get access to the broader set of tools that they’re going to need to solve problems because it can’t all exclusively come from a single provider, right?. We’re all hybrid all the time permanently. From an infrastructure standpoint, I’m going to stay very focused on GCP because I like making promises it keeps. But but it’s but it’s also very nice to have the sort of full ecosystem of tools to bring to bear. 

 

Sasha Kipervarg [00:56:24] Awesome. 

 

Miles Ward [00:56:25] Gents, right, I beat you up for like almost 45 minutes. I’m super impressed. Did a great job for for folks who want to follow up. Like how do they find Ternary? How do they make fun of you at the bird. What’s what’s right next actions. 

 

Sasha Kipervarg [00:56:40] Well, you can just hit us up at our website. It’s http://Ternary.app. And actually, if you want to talk to me, there’s like there’s a link right at the bottom where you can directly schedule my time. I’m happy to get together. And of course, we’d love to continue deepening the relationship with SADA as well, helping with the with the survey and the and the statistics and Lukas – I need  help. Oh, we’d love to provide it. And and of course, Lukas is giving us advice as well. And I am very thankful that he is assisting. 

 

Lukas Karlsson [00:57:18] And for me, you can just follow me on Twitter @lukwam. I mostly talk about our recent COVID testing work for the most part, but. 

 

Miles Ward [00:57:30] Thank you. I enjoy not having COVID as a result of a nation better informed by the hard work you guys are doing. I look forward to reprising this over karaoke. That’s the only reasonable next step as far as… 

 

Lukas Karlsson [00:57:45] I did find video of our last karaoke. 

 

Miles Ward [00:57:50] I remember receiving these DVDs that they had recorded for maybe not the last one, but the one before that. And Elena Gaile came up and she like took them from me and placed them in the trash, like, wouldn’t break. She’s like, these are bad for us and they should go away. I was like, oh, you’re probably right. I was like one karaoke to me. You know, sometimes you got to go hard on karaoke. And I was like… 

 

Lukas Karlsson [00:58:14] A post GCP next karaoke. 

 

Miles Ward [00:58:20] Oh yeah, very good. Sasha what’s for the closing and what’s your go to karaoke song. Come on. 

 

Sasha Kipervarg [00:58:25] Oh, I don’t know, The Aphex Twins, but I don’t know how I sing that. 

 

Miles Ward [00:58:29] You know, just words and beeps coming out of your ears. I love it. I’m down all the karaoke session. 

 

Miles Ward [00:58:35] All right. Bring a one. You bring a one square foot pane of glass and you just look at the window like it’s perfect. Good video. Ah, that’s good video. All right, gents, I appreciate it. Thanks a bunch. 

Sasha Kipervarg [00:58:50] Lukas, see you online.

Google Cloud Logo Icon

THE GCP VS AWS DEBATE

We spoke to dozens of customers who shared their experiences with both cloud providers. The overwhelming trends tell a big story. Download the eBook to learn more.

Solve not just for today but for what's next.

We'll help you harness the immense power of Google Cloud to solve your business challenge and transform the way you work.

Scroll to Top