Solving Application Cloud Waste

A Panel Discussion

Air Date: April 22, 2021

Matt: Alright, well just wanted to thank everyone for joining. Looks like we’ve got quite a few people that have not only signed up, but even within these pandemic coveted days actually joined the webinar! So super excited for everyone that’s here. Thanks for taking the time on Earth Day to talk about a really, really important topic around cloud waste.

Super excited for the panel that we have here today. I’m Matt Provo by the way. I’m the Founder and CEO of StormForge. One of the things today will not be about is a deep dive into what we’re up to necessarily, other than the fact that we absolutely as a part of the technology that we’re building are connecting our solutions and kind of the value that we provide to this concept of reducing cloud waste and over-provisioning. Outside of that, the topic is going to be kind of moderated and driven by me today, but super excited about the panel of guests that we have with us today. So thanks again everyone for joining.

Just to give a little bit of agenda, we’ll spend about 20-25 minutes talking through some questions related to this topic. Each of you can submit questions throughout in the Zoom and we’ll get to some of those across the last 10 or 15 minutes of the panel today, but we will spend the first half, or two thirds, talking through some topical questions that we’ve come up with for this webinar.

As I’ve mentioned, I’ve got Adrian, Cheryl, and Eric on the panel with us today. I could not have asked for a better group, honestly, to get together and talk about cloud waste, the impact of that, and really what we can do about it. Thanks for attending as I said. Your attendance on Earth Day, but Earth Day notwithstanding, does benefit an organization that we’ve partnered with called One Tree Planted, and because we all need to eat, each of you that have attended will be getting a Grubhub complimentary gift card. We’ll follow up with you afterwards and make sure you have access to that. So let’s jump in.

We’ll start by doing a little bit of a popcorn, round robin, whatever you want to call it, between the panelists. We’ll go in the order that you see on the screen and I’ve asked each of them to just spend the two minutes or so talking about their background and then why this topic in particular is interesting to them to participate in the panel. So Adrian, if you don’t mind kicking us off, it’d be great.

Adrian: Yeah, sure. Great to see everyone and thanks for inviting me on the panel. So, Adrian Cockcroft. Currently work for the sustainability of part of Amazon. Most people know me more recently as being at AWS working on cloud architecture and a lot of different things there. Spent quite a bit of time looking at this topic over the years and something, I don’t know seven or eight years ago when I was at Netflix I think, I did a talk on cloud native cost optimization, which is trying to figure out all the basic techniques for optimizing your footprint, which is a lot of getting waste out of it. Nowadays, I think we’re looking at that and then thinking about today being Earth Day, what is the… How do we relate, just general efficiency in the clouds to carbon footprint on top of that, so with a little bit of an extra twist. We can talk more about that a bit later.

Matt: Awesome thanks, Adrian. Thanks again for being here. Cheryl, why don’t you go next.

Cheryl: Absolutely. Thank you so much, Matt. So yeah, my name is Cheryl Hung and I’m the VP of Ecosystem at the Cloud Native Computing Foundation, which is the open source non-profit that hosts Kubernetes, most famously Prometheus Envoy, and many, many other open source projects around cloud native infrastructure. Cloud native from the very beginning, one of the biggest incentives behind it was to be able to use your cloud resources more efficiently, for you to be able to pack your workloads tighter. So I talk to a lot of different end user companies, I run the CNCF end user community, which has more than 140 organizations, and it is something that I hear come up time and time again from people about how to optimize their cloud spending. So in addition to this, I’m also an advisor for a startup called Helio, which kind of builds itself as AirBnB for data centers, but also has a mission to reduce the cloud waste and wasted resources, so I’m particularly interested in today’s topic.

Matt: Awesome, thanks a lot for joining. And Eric last, but not least to do the intro my friend.

Eric: Hi, everybody. My name is Eric Mulartrick. I am a FinOps Lead at Boomi and I have actually been practicing FinOps for about a couple years now. My background is more on the infrastructure side, and we had the typical problem when we moved into AWS that it was a lot more expensive than what we thought it was going to be – why is that? So I took it upon myself to learn FinOps and become the expert here at Boomi. Every day I’m looking for an opportunity to improve our efficiency and I enjoy it. It is my passion.

Matt: Awesome, thanks a lot for joining. Alright, Adrian, we’ll start with you. So this concept of cloud waste has sort of been up for discussion, if you will, and kind of forming in its definition. I know you’ve obviously spent a lot of time thinking about it. Can you just level set for us what cloud waste, at least from your perspective, sort of means from a high level? Then I’ll let the others contribute as well.

Adrian: Sure, so I think the first thing, the thing that’s most different about cloud, is that it’s elastic. The thing… What we got used to in the data center was buying some resource and it was going to sit there in the data center for three years, you pay the depreciation, and you try to guess how much you’ll need and so you never run out. So you tend to over-provision in the data center and your cost model is kind of you commit up front and then you’re depreciating. So when you go to cloud, that whole model changes and you’re paying usually the month after you use something for how much you actually used.

So for example, if you need to store a bunch of stuff on disk, let’s say whatever terabyte of disk space, you need to buy more than a terabyte of disk space, and brought this space because of formatting and things like that, and then you need to make sure you never run out. So if you think you’re going to store a terabyte, you’ll be buying a lot more than a terabyte, but in the cloud you basically just deploy whatever you actually need at the time. If it ever gets to a terabyte, you pay for the terabyte. So there’s this element that it should be elastic. Storage is relatively easy to think about. When you get to compute, you have to remember to turn compute off when you’re not using it. So that’s a big piece of that, and then the question is how efficiently can you scale things up and down and how long does it take? and I think we’ve moved from like virtual machine instances, which can be created in an order of minutes and they tend to stick around for hours or forever, but the minimum sort of usage time you typically saw was hours, and originally they were charged by the hour. Nowadays I think we charged by the minute. Then we came out with containers, which are lighter weight. So when we first introduced containers, there were some surveys looking at, I think it was Datadog or New Relic, were looking at the lifetime of the containers they were monitoring, and they found that they some of these containers were only living for minutes. And that was an interesting sort of viewpoint that you could create things that are very transient. Then if you look at Serverless and the Serverless functions things like AWS Lambda, we’re now charging for that by the millisecond and it’s perfectly reasonable to create an instance of compute that’s going to run for 20 milliseconds to do something and then go away again. So that is something… And then you just pay for that piece of it… So that’s the trend and then the question is, if you look at an overall application architecture, where do you deploy each of these technologies? What’s the right thing for the right thing you’re trying to do? And how do you take advantage of the elasticity so that you never have machines just sitting idle waiting for something to do? I think that’s kind of the most fundamental way of looking at cloud waste. There’s obviously other things, but I think that’s where I’d start.

Matt: Awesome. I love that. Love the characterization there on the transition that we’re seeing as well. Eric, I wanted to ask for your comments on that, as it relates to FinOps in particular of course, but in this transition and how we’re thinking about what used to be sort of reactive in nature and is moving more to proactive in nature, how do we, even from a FinOps standpoint and connecting that to waste, how do we stay ahead of it?

Eric: Sure, so what I like to do is… FinOps is all about three different phases, right? First is the inform phase, so it’s kind of breaking down the usage to understand where your costs are going, how you’re using it within your organization. Then you move into the optimize phase, which is more or less empowering your organization to understand efficiency and to understand where you can make improvements. And then operate is a continuous operation of it, right? you may have one day where you’re focusing on a new product where you can understand how it’s going to be used and coming up with, for example, what the cost per tenant would be, or the cost per some type of unit. And the next day you may be executing a really massive opportunity that would save your organization $300,000 over the next year! But to tie it back to waste, I found that for me enabling engineers to see efficiency is kind of the best way to tackle it. And to do that you can almost think of yourself like sitting in a Toyota Prius, right? you’re driving around in your Prius and the Prius is providing near real-time feedback to you that says, oh hey, if you take your foot off the gas when you’re going down a hill, you’re going to be more efficient. And it gives that data to you instantly. Now if you provide that same data to engineers, they also respond the same way. They say, oh this feature that I released is massively expensive. I didn’t expect it to be like this. Let me go take action to bring it down to the right size or whatnot. So that’s at least what works for me when I take a look at cloud waste in general, from a process perspective.

Matt: Makes sense, and Cheryl, what’s your perspective on not just kind of the definition and how we should think about what is cloud waste, but what are you seeing in the ecosystem across the CNCF and other organizations that are increasingly, I would say, committed to ensuring that this is a part of their architecture and a part of how they’re releasing software and thinking about what they’re building?

Cheryl: Well, I definitely concur with everything that Eric just said and the different phases that he’s seeing because I see the same thing across multiple organizations within the CNCF end user community. Obviously they’re mostly working at the container level, rather than using VMs, but the thinking is very much the same. You provide the dashboards and transparency about cost, about how money is being spent, and then you empower individual engineers to do something about it. So you give them tools, you try and automate what you can, but then you give them tools and say okay, now it’s on you to make this judgment about whether you should be building something this way or a different way. So I can’t really name names because I don’t know if this is privacy or whatever, but I know some particularly large organizations in the tens of thousands and hundreds of thousands of employees for whom they build up teams specifically within FinOps to deal with this problem. And it’s a cross-functional organization, so they work everywhere from executive leadership all the way down to individual engineering efforts. I think that is important to have that broad view over the organization and not just focus on one or the other.

Matt: Yeah, makes total sense and totally understand the privacy stuff, so makes a lot of sense. Eric, I’m curious what your comments would be on how do we take this concept of cloud waste more broadly and apply it, either to ourselves on an individual level or an employee level and this idea around does what I push for as an individual developer, or whoever it is, does that have a broader impact? I’m in a large organization or not. How do I make sure that I understand the impact of cloud waste and what is my kind of contribution to helping remove it?

Eric: Sure, I like to encourage… When I work with engineers, I like to encourage cloud waste and have some empathy with just sitting at your house, right? Typically, I’m gonna turn off my lights when I’m going to bed or when I leave for the day, right? So think about your own resources in that way, right? If I’m going to be leaving for the day, do I need to have these set of resources on spending the company’s money, and of course x exhausting carbon gas, or do I turn it off and minimize the cost of the company as well as the carbon footprint? And what I found is that for the most part, some engineers have different trains of thought. There’s different motivations. One example that I’ve seen from other folks within the FinOps foundation is they’ll compare cloud waste to bottles of beer, where a company or a team is spending 10,000 bottles of beer on nothing, more or less, because they’re leaving the resources on nights and weekends. Whereas you could also compare to a different type of motivation like oh, I really would like to get a Tesla model 3. Well oh my goodness, my team is wasting three Tesla model 3’s per month. I should probably do something about that. But often you’re gonna come back to the dollar bill. At the end of the day, hey this is the company’s money that I’m spending, and if it becomes an astronomical number that could be something that is a way that collectively we can all work together to reduce that number as a company goal. Or you have to look at it and compare it to how much money you’re spending because if you’re going to save 10 cents compared to a million dollars, you’re probably not going to work on that 10 cent one. You’re probably going to focus on some bigger fish, but it all is relative depending on your organization’s usage and spend.

Matt: Makes sense. Adrian, curious on your comments in this transition, you’re talking about more elasticity and different kinds of services. How do developers think about keeping their keeping their applications performing at the level they need to for business reasons or whatever it is, and at the same time continuing to stay kind of connected to this concept of making sure that they’re utilizing and consuming the resources that they need without having to sacrifice on performance, right?

Adrian: Yeah, I think the first thing is to work your intuition is probably wrong. So that’s the first thing to start with. People tend to not really understand what is expensive and what isn’t expensive. So one of the things that I recommend is if you look at your cloud bill every month, how much attribution do you have from that bill, can you take every line item and say this line item is assigned to this team or this developer or this tool or whatever it is, and then it was something that you can either build into your tooling, but a lot of people they get a long way into the it’s sort of messy when they’re getting into cloud and they don’t maintain that attribution. One customer I was working with had about 30% attribution. So about 30% of their bill they knew where it went, and then there was a bit of a crunch on costs and they went okay, how do we save money on this? It’s starting to get big enough that we should be tuning it. The first thing they did was drive that attribution up into the high 90% range, right? So, that’s the most important thing. Like if you can send every manager in your organization every Monday morning you get all things saying this is how much your team spent last week, how much of the total Amazon bill we’re actually attributing to you every week. That was one of the practices we set up like in a decade or so when I was on Netflix. And that meant that everyone could say, and it wasn’t that we were doing chargeback we were just making it visible, and then you could notice how there’s this yeah there’s a thousand node Hadoop cluster they’re looking around that somebody forgot to turn off that they were playing with or whatever and it becomes visible. So one thing is just making that because then you decide that spending a week tuning a lambda function that’s running in the free tier doesn’t make any sense, right? So you’re trying to figure out where is the stuff that scales. Another example I was chatting to a CIO of a company and he said that he’d finally realized that he could turn off QA 3/4 of the time. And he was super happy about the savings, but he was really grumpy that no one had told him this three years earlier when they got into cloud. He said I spent three years wasting a substantial amount. We just saved a load of money and it’s like why wasn’t I saving this over the years, so some of this is getting these practices in early making it so that your QA system is only running, or your development environment basically, is running for whatever it needs to be. Even if you’re using it all the time, it’s 40 hours a week or 50 hours a week, whatever your working week is. There’s 168 hours in the week. You don’t need it on the weekends and a lot of environments, particularly in large enterprise environments, the test environment can actually be as expensive or more expensive than the production environment. For some of the really critical workloads in banking, for example, the payload of what that workload is doing is hugely valuable. The actual implementation of the machines that run it, there’s so many test versions of it there’s three or four different test environments the same size as the production environment. So you get this kind of inversion. Because if you look at something more web scale, Netflix, Twitter, people like that, what they have in test is much smaller than the full scale production. So you kind of have to get your sense of proportion right I think that that’s where i’d start.

Eric: Yeah, just to add on that a little bit Adrian, I mean at the end of the day you have to figure out if you can turn off environments as you were saying, and we’ve been able to turn off essentially all of our development environment, plus we actually turn off our production-like environment on weekends. And that has saved massive amounts of money just because we spent the time to automate the process, as opposed to having you go in there and click on the console and shut down your resources and then turn it back on Monday.

Adrian: Yeah and it has gotten easier over time. We now have the ability to basically take instances and just idle them and bring them back later, so yeah.

Matt: Cheryl, how would you sort of define or put into categories to the different impacts that cloud wastes are having that you’re seeing across the ecosystem and there’s obviously environmental impacts that we’re seeing, and there’s impacts we’re seeing obviously around cost, how would you categorize those sort of thematically around what you’re saying?

Cheryl: I think that’s actually a pretty reasonable summary. So I think most organizations do immediately think of cost, first and foremost before they think of the environmental impact, which makes sense. Cost is something that’s more tangible to a lot of companies, and even though companies quite often do have an environmental department or initiatives, they tend to be a bit separate from this kind of infrastructure work. I do think that what, I think it was Eric, mentioned about automating a lot of this is really important. So for instance, I see a lot of companies who’ve moved from using VMs to moving to using containers with Kubernetes and instantly getting a 2 to 3x resource efficiency improvement from that, which is much easier to handle because it’s Kubernetes doing the work for you rather than you having to manually look at it and make these decisions yourself on an ongoing basis. So I feel like companies often move towards these kinds of newer infrastructures for resource efficiency, or for faster release cycles, or other reasons, but as a nice side effect get the reduced cloud spending and the reduced environmental impact.

Matt: Yeah, I mean just a quick comment, if we think about something like ESG. ESG initiatives are actually making their way into the boardroom and executives and certain companies are now, which I think is fantastic, like a part of their compensation and hitting their goals and achieving what they need to achieve in some cases is connected to reducing or ESG initiatives within these organizations, which I think is again just a fantastic shift that we’re seeing. Eric, I’m curious…

Cheryl: …yeah to make one more comment, yeah 100 agree with that. I think it’s a really exciting development. Bodes well and this is also something that I see through the startup that I advise for, which is called Helio. They really did covered it from the thought that they were obviously the huge public clouds like AWS, Azure, and Google Cloud and so on, but there’s also a lot of data centers sitting around the world that are idling because in previous years and generations, their organizations have over provisioned and no longer make the best use of that data center capacity that they have. So it can be really good for companies if they can rent out that extra space and kind of provide their own model to other companies who would be willing to rent space from others.

Matt: Yeah totally agree with that.

Adrian: Actually I’ve been working on some ESG related topics, so I think it’s good for people to understand a bit more about this if you’d like to stay on that for a little bit…

Matt: Please go and then after that, Eric, don’t let me forget, I need you to answer how did you know what to turn off on the weekend, so we’ll get back to that in just a second….

Adrian: Alright, okay. So ESG’s, Environmental, Social, and Governance, and I wouldn’t say it’s going into the boardroom, it’s actually more coming out of the boardroom. It’s becoming part of the company disclosure that is required by regulations and it’s moving from best practice, which is a lot of large companies are doing these disclosures to being mandated by certain countries, and particularly in Europe. The European Union has been building up more rules about what companies have to disclose and there’s a proposal to make it more mandatory in the US. So this is something that’s sort of top of mind for companies in their quarterly disclosures, on annual disclosures, of what they are doing. They’re supposed to have a report on their environmental impact, what they’re doing from a social point of view, and governance point of view. It includes all kinds of things, not just carbon footprint, but things like do you have any labor related issues? What about your supply chain? There’s a whole bunch of things in there, which are related to this area. And then what we’re seeing though is financial entities are starting to build funds of ESG-oriented funds, so they’re saying we only want to invest in companies that have a good ESG profile. So this is putting an incentive on companies to clean up their act and have a good ESG report, and even to report it. So that’s kind of acting as a pressure for people to clean up their act and some of that involves having a good carbon footprint sort of posture, so that’s kind of the background. And what’s happening is, because this is coming in from the investor community, it’s reaching the board who care about this, which is coming down to the executives, and then it’s coming down into the CIO and the Chief Product Officer, and it’s about your entire product suite. Typically for most organizations, they spend more time moving atoms and molecules around than electrons, so their electricity bill for cloud operations is usually a small percentage, but they still want to figure out how to optimize for that. The real benefit I think of cloud is it helps you transform your product line more quickly to a more sustainable product line. In the same way that we use cloud to speed up digital transformation, we’re using cloud to speed up sustainability transformation. And that’s actually kind of the new area I’m working on more directly. I want to just mention one more thing briefly, which is that there’s an organization called os-climate.org, which is a Linux Foundation non-profit. I represent that organization into Amazon and we’re working on building a data lake that finance companies can use to try and model the impact of climate change on their ESG asset portfolios basically. So whole sort of brain dump on ESG there, but hopefully that was useful.

Matt: Totally relevant and really appreciate it. Your first comment, which is a correction of me and I agree, not going into, but coming out of the boardroom and the resulting impacts both internally with initiatives, but also the external forces around regulations and different initiatives that are driving this, I think is fantastic. Thank you for that.

Alright back to Eric. So you seem to at least have some level of confidence in some cases of what you could turn off over the weekend as a way to kind of take this on, at least in some way shape or form alongside the automation you’re doing, but curious. Just straight curiosity, how’d it started?

Eric: From actually the beginning. So we took essentially what was in a sort of a heritage data center environment and we built it as code within AWS. And as part of that process, we came up with an orchestrated way to power it down in an orchestrated fashion and then bring it back up in an orchestrated fashion. So we started from the beginning and made sure that we incorporated it in each and every part of our product line that we release out there. And it’s to this point, it’s always been within our thought process, at least for non-production systems. Production systems – different story, right? We’re primarily sitting there 24×7, and as Adrian mentioned, being elastic is the way to get around just having resources sit out there, right? If you can grow with your clients, that is much more efficient than over-provisioning and then waiting for your clients to grow into your workloads that you’ve provisioned.

Matt: Makes sense, awesome thanks. So an interesting question that came in actually and we’ll transition just a little bit. I’ll kind of ask one more general question, then we’ll start taking questions from the audience, but interesting question. How do you get product managers on board to connect revenue, to cost, to business value, to how they’re sort of leading efforts on what gets built and how it gets built to a certain extent? Like this concept… I love product managers. I came up in product, but an interesting question.

Eric: I’ll take a stab at that one. First, I think what you want to do is come up with some sort of unit economic or unit metric, such as here’s my cost per tenant, here’s my cost per integration, my cost per gold and record, so on and so forth. And then you can showcase that particular unit economic to say look if we’re going to add this feature it’s going to add x amount of users, right, which is going to offset the particular infrastructure. You could set goals for being flat from a unit economic perspective or even have some sort of growth, but it comes down to like understanding and putting that business context to infrastructure and having north-star metrics. At least that’s the way I’ve dealt with product managers in this regard.

Matt: Okay, Cheryl, you want to add to that?

Cheryl: I have seen in one particular company where at the company level they noticed that their cloud spending was rising faster than their overall revenue. They basically made it a top-down company effort to slow this down over the upcoming years to maintain their profitability. I think having that kind of executive, perhaps mandate is too strong, but very clear direction that this is part of your job, this is part of what is going to help the overall company in terms of sustainability, that made a big difference to the individual product managers, to the engineers, and to everybody working on the product.

Matt: Yeah, makes sense. Adrian, I imagine you’ve worked alongside and worked with your fair share of product people, heads of product, etc both internally and maybe even externally. Any comments on product management in particular and connecting that to revenue and business value and ultimately cloud waste reduction?

Adrian: Yeah, in my career I’ve actually been in product at various times as well and worked on that side. I’ve got a very strange mixed up resume. So I think that the real thing… One example I had and this goes back to about 20-25 something years ago, I was at eBay in the early 2000s, almost 20 years ago, we used to charge things… We used to charge back to projects for the cost of getting them in production. Like how many machines do we need to buy to launch this feature? and then we switched it to this is how many machine weeks we’re charging you for running this feature. Alright? so we went to an activity-based costing kind of model, rather than a launch based costing. It’s a very fundamental thing. There were enormous arguments, just enormous internal arguments around it launching this, but when we did it, the reason we did it was we were out of data center space. We were running in data centers that kept filling up and we couldn’t launch new features because they were full. So we had to create an incentive for people to clean up their data centers. And we immediately found all kinds of features that were still running that weren’t adding value. People said okay yeah, let’s just reduce this. We had project proposals to take things offline. We freed up hundreds of machines really quickly, and effectively I think we lived off of the spare fat in the data center for a year or more after that before we actually got to an efficient setup. So this was like 2006 5 2006, but I think that that’s kind of the way to think about it. You have to have some kind of chargeback, which lets the product manager have an ROI. What are we spending on this feature? What is the return? And does that make sense? If something is marginal, then maybe tuning the application and trying to make it more efficient. You can take something from a net loss to a net profit by just tuning the… If there’s a significant amount of costs baked into the business model for it. So it really comes down to having individual little business unit chargeback models.

Cheryl: To add to what Adrian just said, I’m having another example in mind where a company got quite obsessed with the idea of optimizing the hardware in their data centers with the idea that it would get them a couple of percent of better efficiency. And it took quite a while for them to realize that maybe you’ll get an extra 10%, but the hardware is very limited in terms of how much extra capacity you can squeeze out of water calling or packing servers more tightly together. Whereas changing the cloud infrastructure and changing things in software, you can get 10x improvements in efficiency. So a little bit of a warning story.

Eric: Yeah, that’s really where the benefits are gonna come in. It’s the re-architecture, right? I can only right-size so much, but if you really want to take a stab at it, it can be switching from EC2 into containers, to start packing the EKS clusters full, or it could be re-architecting the application to really start growing with your clients versus over-provisioning.

Matt: Yeah, makes sense. Alright Eric, I’m coming back at you my friend. What is this thing called FinOps that’s growing and is getting greater adoption and how do you connect the world of FinOps to reducing cloud waste and cost?

Eric: At the end of the day, FinOps is…

Matt: This is an alley-oop for you, my friend. This is like an alley-oop for FinOps.

Eric: FinOps is something that, I mean you could basically look at it as shorthand for cloud financial management, so a term that has existed a little bit more at least in your past. But FinOps essentially was birthed out of the Cloud Ability, I believe it was, from J.R. Storment. J.R. Storment is the founder of the FinOps foundation. He was one of the founders of Cloud Ability as well, but what he was seeing often was that a lot of customers were asking the same thing. Is there anywhere we can talk to have best practice discussions and do you work on the same problems that we’re seeing within our organizations? and J.R. birthed it about a couple years ago and we’ve gone from… I remember two years ago joining the sessions, we had maybe 30 people on a call? and now the past call that we had a couple weeks ago, we had over 500, or close to 500 excuse me, which was incredible! But more or less, the FinOps foundation is a spot where we can talk about those gnarly problems that everyone is having, as well as establishing a framework for all working on the same problem together. And like I said it ties back to cloud financial management, right? it’s got three phases. It’s the inform, the optimize, the operate, which all dovetails into what cloud financial management is, but it’s establishing these best practices within each organization, so we can all think a certain way and change the culture within an organization to start looking for that cloud waste and understand where my costs and usage is going.

Cheryl: I have a question for Eric actually. FinOps feels to me… I’m very excited by FinOps, right? Recently I posted about my top 10 trends and predictions for 2021 and I named FinOps as one of those predictions. FinOps to me feels a little bit like the way DevOps felt maybe five years ago, where it was about best practices, and human culture, and the kind of tooling wasn’t quite there yet. So are there things that you see as upcoming spaces where there could be more tooling or… Yeah, what do you see coming next?

Eric: Yeah, yes and no. I basically… You can look at tooling in a couple different ways, right? Do I use vendor native tooling like AWS cost explorer, right? Or do I buy something potentially that will give me those metrics and data and visualizations that I’m looking for? As you move into containers, there are certainly different, I mean Cloud Ability for example supports containers, but there are like cube costs. I think it is a piece of software that you can buy that will help you within Kubernetes tell what or who’s using what. But I think that plus some open source utilities, there’s also, just to throw it out there, Cloud Custodian, which is something that came out of Capital One, is another way to kind of lay down a level of governance on top of your organization. So if I have resources that come out that don’t meet my tag requirement, I will go destroy those resources or shut off those resources. But I think at the end of the day it merges together with DevOps, right? because you’re educating your engineers and your developers about how they are using the cloud and leveraging DevOps tools maybe you have it as part of the pipeline, where you can see if it’s right size from the beginning per se, right? So just that it’s a mindset that is converging together between DevOps and FinOps, in my eyes.

Cheryl: Awesome, yeah. Thank you for that.

Matt: So we are, out of respect for everyone’s time, we’ve got about three or four minutes left. I’ve tried to take some of the questions that have come in as well from the audience. Adrian, I’d love for you to answer Nikki’s question live just real quick, if you don’t mind. Just reminding the group of the OS initiative that you mentioned. Then if you can be the first to close us out. Just each of you end on we’ve talked about what is cloud waste, what is the impact of it, and it’s more than fitting to then kind of close out with what can I or what can we, as an individual or as a company, kind of do about it? So if you can just comment on those things, we’ll wrap up.

Adrian: Yeah, I put a note in the chat session saying it was os-climate.org. That’s the organization I was talking about. We’re currently working towards an initial release of the open source project, so there’s nothing quite there in GitHub yet, but there will be a basically a data lake that lets you look at physical and market risk and figure out who owns properties and things like that, so there’s a whole lot of stuff there. but on going back to the topic, I think one of the things that people can obviously do is optimize for cost. And it’s pretty easy to justify that you’re going to save money next month in the cloud if you can cost optimize. Almost always that will also reduce your carbon footprint. They are very closely linked. There’s a few places where you can do a little bit more of an optimization, and there’s a few other things to think about. Some of this is like in security, there’s stuff that’s the customer responsibility and some that’s the cloud provider responsibility. There’s the same thing for carbon footprint, right? There are things you can do and there’s things the cloud providers are doing. So right now Amazon overall is the biggest buyer of green energy in the world, and the biggest in Europe, and earlier this week we announced two gigawatts of 1.9 gigawatts of additional purchases. So it’s a… we’re spending a lot of money to get all of Amazon to be running on completely green energy. It’s more complex than that, but that’s kind of one of the things we’re doing is investing on our side to make sure that we’re just taking care of these things for you. So that’s one piece.

If you think about the things that don’t completely line up with cost, one of them that you can do is think about all those cron jobs that you run on top of the hour. Think about how many machines across AWS wake up on the hour, or at midnight, and hit something. They fire off a lambda function. They call they do something to s3. They hit our control planes and we have big spikes in traffic on the hour, so one thing you can do is actually called cron jittering. You just make it so that your cron jobs don’t all run at that peak time. If you’re doing a daily roll up, maybe run it in the early afternoon when there’s lots of spare solar power and generally the grid is running greener. So the time of day matters. So if you want to ask the question, what’s the carbon footprint of my workload, it turns out the answer is extremely complicated. If you just try and measure the power consumption of your laptop when it’s running something, it will be all over the place. Like you’re running virus scanning at the same time? There’s a whole bunch of things happening that are really hard to one pick, but directionally in almost all cases, just optimizing for cost is going to be a double win, and then there’s a few extra things you can do on top where you can maybe do a little bit more optimization. But some of those are going to be taken care of by the cloud provider over time, so it’s more like thinking about just some incremental things on top.

Matt: Awesome. Eric, I’m going to go to you next, but before you give your answer on what we can do about it, I’m going to rapid fire you a couple quick questions that’ll be quick answers from the audience. Can you confirm your opinion that cloudcheckr is an example of a FinOps tool?

Eric: Yeah, definitely. Yeah, cloudcheckr is absolutely an example of one.

Matt: Awesome and then the same person would like you to repeat your FinOps something inform operator?

Eric: Oh yeah, I’ll do it better. Let me share a link to the FinOps Foundation and you can take a look at the website. it’s free to join. Everyone’s welcome. So I will, unless you’re an actual like provider of software, there’s an avenue to become a member as well, but it is all on the FinOps page. Let me drop this into the chat for everybody to see.

Matt: Awesome, and then what can individuals or companies do to help reduce their cloud waste?

Eric: I would start with understanding your cloud. As simple as that sounds, it’s more or less coming up with a tagging strategy that meets your cloud footprint, so you can understand where your costs are going. Then once you have that understanding, you can start showing it back, as Adrian mentioned, to your engineering teams and your leadership to understand where you’re spending money in the cloud and how you’re using it. Then it’s the Prius effect after that, right? Enabling that efficiency to understand your efficiency.

Matt: Yeah, I think a few people liked how you connected it to Teslas, so that was… people like that. Nice work there. Alright, Cheryl, bring us home. What can people or companies do to help reduce their cloud waste?

Cheryl: Yes I have two thoughts on this. One comes from a Summit that I attended in Geneva run by the UN a few years back about what is the best way that we can reduce the environmental impact of infrastructure as a whole IT infrastructure, and my best answer at the time was to move to public cloud. If you are not yet on it, which seems obvious because we’ve all been working in cloud for a while, but there’s still a lot of companies out there who are running their own hardware and they’re never going to have the scale or the ability to do what Adrian described and offset all of the carbon footprint that they’re using. So if you’re not on cloud you should move to public cloud and it will reduce your environmental impact.

Then the second part actually is closely related to what Eric mentioned. I really admire the FinOps foundation and what J.R. Storment has done for that. I think it’s a really fantastic way to find out from other people what they’re doing. You’re not alone in this, everybody’s trying to figure out the same best practices all at the same time. So go out there meet the community and learn from each other and share what you’re doing.

Matt: Cheryl another question for a takeaway, is it best practice to lift and shift, or is it best practice to re-architect and shift?

Cheryl: We can talk about that in another session sometime. That’s a long question. Long answer to that.

Matt: Well I just want to say thank you to Cheryl, to Adrian, and to Eric for joining a super fun conversation. We at StormForge – and I’ll just close with this so that my marketing team doesn’t get mad at me – we have a pretty awesome pledge that as a company that we are committing to remove $100 million dollars worth of cloud waste in 2021. We’d love for anyone attending or interested to sign up for the same pledge. We’re a smaller organization, so to remove 100 million for an organization of our size is a pretty bullish initiative, but we’re super, super excited about it and we’d love for anyone who would like to go to the link below here and sign the pledge. We give you details about things that you can do to be a part of erasing cloud waste in your environment.

With that I just wanna say thanks. I know we’re a couple minutes over. Most everyone stayed the entire time, so this must have been at least somewhat interesting for people. Appreciate everyone’s contributions. Again, to Adrian, Cheryl, and Eric, thanks so much for your input and your contributions. And thank you so much everyone for joining.

Cheryl: Thank you so much, Matt.