 All right, very well So good afternoon everybody. Thank you for coming and attend this session We are Eric and Massimo and we are part of the team Responsible for the product strategy inside the management view at Red Hat and today We are going to talk about the financials behind the private cloud and Specifically behind the private cloud based on OpenStack Before showing you some numbers I'd like to start giving you Some details about the background of this project because we started it as an internal research so no plan to share what we were going to to find out and We started this research to find the answer to two big question. The first is Is OpenStack more cost-effective to build a private cloud if we compare it with other Non-OpenStack based solution and the second question is okay Convenient or not. We choose OpenStack deployed it. Okay. What are the first thing to look at? What are my? Priorities if I want to take out the most from my private cloud so we run through a number of TCO models that we found on the market but To be fair, they were not good enough to provide the kind of detail that we were looking for that We're not transparent enough for our purpose. So we decided to build our own model and What you will see is very different from what you may be used to because it's not the standard marketing tool Conceived and built just to prove a couple of points that are convenient for us as a vendor but It's been bill has been billed as a tool first and to be consumed by IT decision-makers and We put together a tool able to provide support for financial conversation and within the company and with the line of business and During the process of building this model We requested a number of peer review from IT guys and financial guys Across a number of companies Basically, we followed the exact same process the exact same methodology that the major analysis firms does to do this kind of research To be as objective as we could wherever we could we relied on start on standard market analysis data But for those kind of numbers for those kind of information that they are not covered by traditional Research we had to do a number of assumption. So Something may be wrong, but if he is it is just for tourism So or we did the wrong assumption at the beginning or we used outdated numbers But the good news is that the model is completely transparent So it's very easy to fix both these two dimensions and to improve this model over time the important thing to know is that we tried to build the most comprehensive the the most Objective and the most real TCO model that we could to describe a private cloud and we built it to Work as a scorecard. So to help doing comparisons using your own numbers Long story short the model right now is able to provide numbers to be crunched by financial guys Has been designed by well IT guys who knows things like that employees and technologies has different cycles within the company and Because we believe that Tools are the right way to find out answers. We build it to support organization and Ourself at the first time into ongoing decision processes So as I mentioned before we started this research as an internal activity Absolutely no plan to share the results but in the process of building this model we found out a couple of Remarkable thing very interesting things so we decide to submit this session to the summit. We are very happy that has been accepted and I Don't want to waste any more time with an introduction. I don't want to go straight to what we find out and the first thing is that good news Open stock is the least expensive technology on the market if you want to build a private cloud and That's a big result not really surprising But because no one tried before to build a model able to describe a private cloud from a financial standpoint We are more than happy to have numbers that can prove this point But the real amazing thing of this result is that is not related to a single situation Is not related to a bunch of use cases that we choose wisely to have this result because more than being The least expensive technology on the market no matter how fast you grow no matter how slow you grow no matter The the adoption curve that you will see the model tells that there is no single situation Where open stock is more expensive if we compare it with known open stock based solutions Okay, it's time to give you a little bit of background on how we achieved those results and We started considering a number of different growth cures Why because we try to understand if the result that we had from the model where consistent and unrelated to the adoption curve and Those cures are all different because describe Different kind of adoption strategy that an organization may choose and how successful those strategies are Those cures are different because are influenced by a number of external and internal factors For example, you may see organization going for an all-in strategy Moving all their application on a private cloud On the opposite you may see organization betting on a single relevant pilot project that if successful is able to drive a huge amount of confidence within the line of business and Because there is this independence that that that we so between the result relative to competitors and the growth cure you may see Numbers that are slightly different from what you see for your organization But the point is that the model remains valid So because there is this independence and so no matters the growth cure of the results remains valid We decided just for this specific representation to use what we think is the most common growth cure So the exponential cure and if you attended to the keynote session You may have heard about a number of use cases of open-stack users that showed up numbers that support this kind of statement So We say that the model is made to do comparison So comparison are the way we use to answer the two question that I explained before and For the first question we made a number of assumptions. The first is that You need and you want to deploy a private cloud So your choice is between open-stack and a number of other technologies a number of known open-stack based technologies That's why to answer the first question So if open-stack is the most cost-efficient we did a comparison between open-stack as a foundational technology For a private cloud and other technologies to answer the second question We did another kind of comparison that is Within the buying model that you may choose to adopt open-stack. So basically open-stack versus open-stack But what we compare actually in the model We compare the three main cost items inside a private cloud They can be divided basically in what you buy and whom you work with So what you buy we are talking about the software and the hardware and whom you work with we are talking about the people and Those three big cost items are made of a number of different dimensions For example for the software we consider subscription and licensee cost depending on the cloud software Of course, and for the people we consider things like Server and storage administrator Networking and system engineers. We consider their managers The onboarding costs and the training all those kind of things for the hardware We consider all the things that you need to buy so server storage networking and All the money that you are going to spend to renew those technology over time What happens is that the model itself is able to address each one of these cost items over time in a different way For example, if you want if you want to work on the software cost The model compares a number of different cloud technologies to find out which is the most efficient And if you want to address the people's cost the model simulate the introduction of it automation in your infrastructure In the end if you want to act on the other cost the model simulated the increase in VM density per physical host So what are we going to compare? We are not quoting explicitly any specific competitor because it's not a nice thing to do but The matter of fact that we all know that all the competing technologies offer a certain degree of automation and orchestration capabilities and Because we need to make an apples to apples comparison We can't use a plain open-stock distribution Because we all know that open stack has a lot of potential for automation But right now it doesn't really provide those kind of capabilities that the competitors have in their own Products in their own bundles so What are we using here is open stack plus a cloud management platform able to provide those kind of capabilities and Well because we are at that the numbers that we put inside comes from RHCI right out cloud infrastructure Which offers both open stack and cloud form as a P and as a CMP Now let me spend just few moments on the notion of CMP because cloud management platform is a terminology that Gartner introduced to identify a family of products intended to augment infrastructure provider such hypervisors Infrastructure as a service cloud managers like open stock itself even platinum as a service and The kind of capabilities that you may expect to have a CMP to provide to your infrastructure are things like chargeback orchestration service catalog external crowd brokering whatever and Another thing that you should expect the CMP to be able to do is to provide all those capabilities Across a number of engines on premises and a number of different public cloud providers Enabling what the market now calls a hybrid cloud So in the end we are talking about a piece of software you may expect to have into production in the last stages the last maturity stages of your infrastructure and And That's lead us to the point where we should remember that implementing a private cloud is a journey and It may take years to reach a certain degree of maturity or even full maturity That's not something that we are saying is not something that Red Hat says is something that a Large number of analysis firms are saying so far Let me clarify just for a moment this thing you can deploy infrastructure as a service in let's say six months But you'll still need years to reach different stages of maturity Because for example, you will need time to align the skills of your people to the new technology You will need time to align your internal processes your business processes to the service model You will need time for example to identify and deploy a number of tools able to provide new capabilities to the private cloud or to augment it like Introduce the CNP and all those things while you are still driving the adoption of a new infrastructure that you put in place That's the reason why we are showing this comparison on a six years time span That's one reason the other reason is that we are showing the cost and how this does cost changes over time In fact, we developed the model on a ten years time span Which is way longer than how your current implementation may look like but we did that to be sure that there were no weird gotchas over time so couple of things that you may need to know to read in the proper way this comparison the first is that because we choose to use for this specific representation the Exponential growth cure you will see the number of VMs doubling here over here and another very important thing is that The metric that is representing the cost here is a unitary metric. So it's cost per VM This is a very relevant thing, but we have an entire section about that So we can just go through and finish this boring long introduction and so I show you graphs and On top of the graphs numbers Those are the numbers who actually Support the two statements that we did at the beginning and because this is a crucial point of this project and a crucial point of this Presentation I'd like Eric to spend a few or even a lot more words on the background of this project Thanks. Thanks, Massimo One thing you can see here is looking at some of the difference in the cost So in that first year maybe five hundred dollars cheaper per VM As you're further along in different maturity that increases maybe seven hundred fifty dollars If you are looking to convince your CIO and say great fantastic. We have this project This is the technology I want one of the question he or she is probably going to ask is all right So it's a little bit cheaper. Is that cost going justified for the risk? So one thing to think about is you know, you can see how that increases over time So if we look at the aggregate costs, you know, these costs get fast because we're growing fast And you can start to see look at all right while it's still growing fast But those curves look kind of close together Tell me more about that. So let's take a look at what we have for just that difference that gap between So because the non-open-stack technology even with the automation orchestration everything else is more expensive The difference between those two increases over time. So we start to see really big numbers popping up pretty quickly and you have to remember that these numbers are based off that relatively slow start of the exponential curve My guess is there probably a lot of you in the room who may even have more Then we have more VMs out there than the number that we have at the end So if you have, you know, that means that you're reaching those savings even faster One thing another way to think about it though is is we look at the not only the cost But what is the total cost of running this over time? so the majority of this cost is in software cost there's a little bit in Hardware, but we'll talk about that later So the majority of this is in software cost. So if you think about this as You get Different amounts of discounts from your from your different vendors So not surprisingly our competitors did not want to give us their discount schedules So to do an apples to apples comparison. These are all this prices so if you think about it and you think about the the Numbers that you have the discounts you get from your competitors just than by your discount It'll be pretty close to what you see but even with those discounts. It really doesn't matter what those discounts level levels are pretty quickly you can see is if you get to a Private cloud of any real size You find that the open stack with a cloud management platform for the automation Will save you millions of dollars over anything else that's out there on the market so we're gonna hold them to the end but We'll keep going thanks in Okay, now what what Eric just explained is basically the TCO behind the private cloud and the Remarkable amount of money that you can save choosing open stack plus a CMP But if you remember the second question was okay, but there is any way for me to save even more money there is any way to increase those efficiency and get out even more from what I have bought and Luckily the answer of the model is yes, so the the the question now becomes yes, but how and What the model points out is that? Automation introducing automation has the highest impact on cost by and far than everything else So automation that as we remember impact directly on people cost is actually the first thing to look at to Lower and be more efficient from a constant point on your private cloud All right, so remember we said at the beginning what We're using a lot of industry data so we can be as pragmatic and practical and objective as possible one of those pieces is the level of automation and the impact it has in the number of OS instances so VMs plus physical per sysitman So it's an organization by name of computer economics does a lot of this kind of stuff They do staffing ratios and other things like that and they find that the average Organization with the average amount of automation supports about 53 OS instances per admin They also find that the average organization with a high level of Automation supports almost double that at 100 and there are some really big numbers even well above and beyond that So, you know as we look at our exponential growth as an organization you have a couple options, right? So next year you're supporting twice the volume that you are today You really don't want to kill your admins not only because that's impolite, but there are some other reasons as well But so you have some options you can either double your staff which good luck with that Or you can introduce something like automation to help you know help that happen So but this is a you know, this is a talk on the on the numbers and the cost benefit so You know the automation in addition to making people's lives easy Is only effective in terms of the conversation that we're having if it has an economic benefit So in order to look at that we looked at a number of different ways that you can purchase consume procure grow your own OpenStack cloud so we looked at the upstream pulling it down yourself The a commercial distribution and a commercial distribution with a cmp So, you know you may be asking yourself and I hope you're asking yourself because we asked ourselves these questions You know, why do I want to pay for an expensive cmp? Why do I want to use a commercial distribution? I've got a lot of really smart people they can support the code they can support They can support the cloud well What the model showed us was that? Despite the additional cost of using a commercial distribution it ended up being less expensive than using the pure upstream and It was mostly because you end up with a commercial distribution just supporting your cloud Rather than supporting your cloud and the code, right? You need more people to do that and then with the cmp the level of automation and orchestration and and all the things that Muscle talked about earlier that high level of automation drives the number of OS instances per admin even higher reducing that cost so despite the additional cost of the cmp Above either of the other options. It ends up being the least expensive as your cloud grows So where did what do those actually look like now? So this is called a waterfall chart You read left to right. We're starting with rel OSP, which is our commercial distribution of Open stack. This is the loaded cost of a VM So bringing in all those costs that we talked about before all the hardware the people the software, etc As you read read to the right you add in the red cost The red makes it higher green makes it lower and then you get to those change and get to our R HCI Which is red hat cloud infrastructure that Osmo mentioned it's the combination of open stack plus our cmp cloud forms and Even with that additional automation cost so their software cost and the difference between the two is relatively small It's about 53 bucks but the tremendous amount of savings in the amount of Additional people that you don't have to hire as your cloud grows ends up being By and far it's huge compared to the the little cost of the software So as you're growing rapidly you end up having to hire fewer people to be able to stand with the growth So there's a there is an additional cost for the software the ROI is kind of obscene if you just look at the numbers on that side And this holds not only in early, but also in a mature cloud. So we're showing your five on the chart So that's exactly the reason why we recently announced the acquisition of the popularity automation tool Ansible But the question now is why Ansible and not other technologies on the market in the same segment. Well Couple of reasons, but just to stick on the conversation. We are having today Probably the main reason is that Ansible is simple to use simple to use means that If you think about the playbook, they are basically human readable that means that you are lowering the cost for the learning and Ansible is agent less and based on SSH That means that you're lowering the cost to integrate the technology within the existing infrastructure Now if you remember the two comparison we say that we were looking at unitary costs So VMS cost and this is a big thing because one of the big take-home of this project is that of course Is extremely relevant to track the total cloud cost But the real metric to look at if you want to show how good you are into providing Efficiency and a good service a scale is VMS cost. Why very simple VMS costs are very good to representing the how cost changes over time and Is a very good metric to represent how good you are into providing efficiency a scale and Finally if you act as a service provider even internally the kind of numbers that it provides are closer to what your customer expects That's another big take home here All right, so a couple things to think about when you're talking about your your units So what do you what are you looking at? So one costs are absolutely dependent on timing People come and go at different rates Hardware you have to buy hardware for growth you have to buy hardware for recess cycles even on those refresh cycles you have to buy up more hardware for growth and Software if it's a license and maintenance versus subscription all these things come together can hit at different times So the the lesson here in this particular is if you offer the exact same service at the exact same quality At the exact same quantity next year your prices might not be the same as they are this year So the other one is if we take a look at the total costs Muslim and I have spent a lot of time with this data We you know we've talked a lot of people about it. We're thinking a lot about it I can't tell if we're getting any better from this Right, I said, okay fantastic now. We're spending a lot of money Okay, is this great? So if we look at the VM costs we can actually tell that we're getting a lot better Right so that that huge curve that we saw before just looking at the total costs I can't see that between year one and year three my cost per VM have been cut in half I can't see these level of improvements So, you know being able to figure out what your unit is that you want to measure and I suggest taking a look at the VM costs will do a significantly better job of helping you Go back to your organizations and say not only are we doing great things But it's helping us change the cost of what we provide you over here and Implications of what Eric just explained is that you should expect your total cloud cost to increase While you are seeing your VMs cost decreasing Why very simple because the total cloud cost is a great indicator on About how good you are into onboarding new people into your infrastructure on the other side We have the VMs cost which is a great indicator to show how good you are into providing efficiency from us from a constant point Inside your infrastructure So there's something a little bit strange about that interaction. So not only we showing different directions But that there's a direct causal fact Lowering your VM costs will increase your usage Larger than you reduce it. So you will increase your total costs by reducing your unit costs So that may sound a little bit counterintuitive But if you look at it just kind of a different way You lower the costs people get more excited about it more people want to use it you grow and the total cost is even more So that growth and cost can be a great thing With clouds and servers and VMs we can actually be even a bit more prescriptive about it So we looked at a bunch of the data from IBM or sorry from IDC about VMs and server costs and things over the years and what we found is that a 1% reduction in cost ends up leading to about a two and a quarter percent increase in usage Which then causes about a one point two percent increase in Total costs, so this has been the numbers are about that range have been the way for and since Give or take 1970 while before we had where we had clouds Now we explained that To lower the cost rate to the software you must choose the most efficient cloud software You can find on the market on the other side We explained that to act on the people's cost you may introduce IT automation first So what remains out the cost of the hardware? What the model says surprisingly is that increasing the VM density on physical servers doesn't have this big impact on the cost Well, we are not saying that is not a relevant amount of money Especially if you reach a certain scale the lesson here is that if you are focusing on high impact cost items Hardware is not the place where you should start you can easily care less about how much you are paying for your hardware All right, so I was surprised You know hardware cost is a really easy one to point to and it made You know my thinking fantastic. Hey if we can double the number of VMs per server then we get to buy a lot less hardware Ended up. It didn't really have a big impact So this chart as you can see all these lines are very close together So what we're looking at is those unit costs those pertin those VM costs per year and we're looking at four different densities So the most expensive is 10 VMs per server the least expensive is the highest density at 30 so if you take a look at somewhere around an average density of say 15 and You double it to 30. You're talking about a savings of about $350 per VM per year And that's before all the costs of software to be able to get that before all of your effort and staff time and everything else to be able to get all these things to happen now Don't get me wrong every single dollar every single yen every single year you can save your organization is important And is a good thing But don't start with hardware density the you know the that $350 even before your Before those additional expenses and effort to come in are significantly less than the thousand bucks or more per VM you can get by Expanding and growing your capabilities and automation Now let me recap for a minute with what we saw today So we saw that if you want to act on the software cost You may want to choose the most efficient cloud software on the market And well you did well choosing open stack, but now that you have deployed it. The point is I Have to Put in place a number of recommendation to get out the most from what I bought and those recommendations are First that you should focus on IT automation that impacts on people's cost Because on on the other side we have VMs cost Which has a surprisingly low impact on the total cost the other thing that we learned is that The metric to focus on to show how good you are into providing the service is the VMs cost The third thing that we learned is that you should expect your total cloud cost to rise While you see your VMs cloud cost to lower and that's because and high Total cloud cost means that you have a lot more people using your service, which frankly speaking is a good problem to have All right, so we talked about a little bit about the model and how it works and some of the results We'll talk a little bit about what we can do in the knobs and the tweaks on the model itself You know we mentioned we made a number of assumptions. We might try to make as few as possible We tried as much as possible to use Data from the industry So here are you know here are some of the assumptions and data that we used and probably a pretty good general classification One thing about TCOs, so I Probably I know you don't really care how much work this was to do the TCO But we put a lot of work into making it really good for you, right and we But you know TCOs like all the financial models are really specific to the individual organizations if you are a if you're an auto manufacturer, you're gonna have a different cost model than if you're a Retail organization and if you're a retail organization in Tokyo It's gonna be a different cost model than a retail organization in Toledo whether that's Spain or Ohio, right? so All of these things have to be able to be tweaked the second piece is TCO. This is you know This is just cost What we really want to get to above and beyond with this is what's the value of this stuff, right? How does this help our organizations? What is the value we are providing to our organizations and TCO is a fundamental piece to help us get there? but We'd like to see more so We got a question for you We're thinking about open sourcing this We're looking for help. We're looking for ideas and we think that if we work together We can take the basis of what we have and actually build it into What if we had a model to look at the value? this is providing to our organizations and you can be able to tune it to exactly your numbers and you know poke at our assumptions tell us we're wrong and all kinds of other things that That we've worked together, you know just some ideas We tossed around that we'd love to see somebody do is what if this is rather than just looking at the infrastructure what if this is looking at specific, you know full application stacks or You know per industry data or all kinds of things. There's lots of places this could go Okay, I think is all for today at least on our side. I suspect we are a bit short on time to answer for question Okay, but we'll be more than happy if you want to reach us any time We will be more than happy and we'll hang out right here until we get kicked out if you have questions So, thanks everybody. Thank you. Thank you for coming