This is another session from Cloudcamp that was held in Toronto on April 6, 2010. (Find the article about the first session, with links to the CloudCamp sound files and some presentations, here.) The session was facilitated by Dave Nielsen and discussed the ROI of cloud computing. The following questions were raised:

What is cloud computing anyway?

How do we measure the ROI of migration to the cloud?

What’s the cost of intangible benefits?

When does it make sense for a startup to use the cloud?

Not all of them got answered but some certainly did, and the answers given were very interesting.

(Note: I did not catch the names of all the people who answered questions, so if one of them is you, please let me know so I can acknowledge it properly! – Tania)

Dave Nielsen: I’ve been saying, “Here’s what I think cloud computing is,” over and over and over again and it’s changed a little bit every time, but actually hasn’t changed much at all in the last like 10 times I’ve done it. But it still could. I’m hoping to get to the 80/20 rule where I come up with 20 percent of what is the main thing of cloud computing and 80 percent of the people agree. But basically, here it is: so you know, you guys know the triangle, the pyramid, cloud computing, Infrastructure as a Service, Platform as a Service, Software as a Service, right? Right here. This is a very, very simple, like over-simplified definition of cloud computing, types of cloud computing.

So Infrastructure as a Service is really providing a service to IT folks. And Platform as a Service is really providing a service to developers where they can put their code. And then Software as a Service is providing a service to business users who don’t want to have to set up anything, don’t want to have to install software on their desktop, right? That was basically the three types of cloud computing but if you don’t know who you’re talking to and they ask you what cloud computing is and you don’t know what type of person they are, or you simply ask yourself, what do all these things have in common, it turns out they really have, in my opinion, three +1 things in common. And the first one is super obvious. What do you think that is?

-              Hosted by somebody else?

-              As a service, all right? So somebody is hosting it for you, right? That’s the really, really obvious one; you shouldn’t have to actually touch any hardware. Okay, so that’s the first thing. So managed by someone else. First definition of cloud computing. Number two, you shouldn’t have to call somebody up and order something that then gets delivered to you that you then touch. In fact, nobody should have to go touch something in order for you to use it. It should be on demand or self service. You want it, you go to a website, fill out a form, you get it. Okay? That’s the second part of cloud computing. And then the third part is, is that you need to know how much you’re using. There should be some sort of metric or reporting that you’re getting that tells you how much you’re using so you can decide, do you want to use more or less? Okay? And all three of these fit that requirement. And then the +1 that people keep throwing at me every time I tell them those three, is they say that it should also be elastic. I don’t think we need to add the third and fourth one in there because I think when you’re on demand and you can choose how much you’re using, that covers elastic. But nonetheless I’ll put it in there as like the +1.

You can automate instantiating more or less, so it’s offered by somebody else. There’s an API so you can automate it and then the metering part’s pretty obvious, you pay by the hour. Or if it was storage part you’d pay per megabyte. So there’s some reporting you can do on how much you’re using, right? And there’s bandwidth too. So you know exactly how much you’re using and you can choose to use more or less. Platform as a Service, again, managed by somebody else. The unmanned part is pretty obvious because all you do is put your code up there and it runs, so it’s not like you have to wait for anybody to set that up for you. And the reporting… So the metering part is usually sort of on a… It’s like a per CPU thing, that is usually how they’re charging, okay? It’s like not per hour it’s more like how much…

Because of Infrastructure as a Service, your server has a limitation on how much one server can compute. But in Platform as a Service they don’t have that same limitation, so what they do is they simply say, “Well, we don’t have the idea of instances so instead, we’re going to simply charge you by how many clock cycles you use.” So it’s a slightly more efficient model and it’s metered differently. And in Software as a Service, same thing; managed by somebody else, usually it’s multi-tenant so you go sign up for it and you get another instance. You know, another account, basically. And then the reporting, the metering is usually by the user, per month. That’s usually how they charge. So again, you’ve got some sort of granularity, right? So that’s basically those three words. And then if you want to be a little extra specific you can add in elastic.

And then at the end of the day, if you’re trying to come up with a marketing term for cloud computing or a marketing definition, marketing folks usually want three words, right? that’s kind of what they’re looking for because if you say something quickly to people, that’s about as much as they can usually remember. And that’s why if you ever want to see a more concise definition that’s more specific and more detailed, look at the NIST definition, which is the National Institute for the Government of the United States. And they have done a really good job of defining it, but it’s not really something you’re going to repeat to other people.

It’s like five steps and it’s kind of more complicated. It’s very similar, by the way. But one of the things that really made me happy was I was talking at CloudCamp Chicago and the people who usually give me the most pushback on these definitions are the companies who provide hardware. Because they want to be cloud, right? But I was talking to this guy and I wish I could remember his name, I’ll remember later. He’s from EMC and he came up… He said, “My definition of cloud computing, it’s three things and you can map them directly.” It was really nice and so I was like wow, even a hardware vendor has three things that are very similar. So when you’re talking about what is cloud computing, if you want to just be very generic about it, there’s three words for you.

-              And what are the three words again? Managed…

-              Managed, on demand and metered. MOM (laughter).

-              Well, and that’s the point, right? So whether it is intangible or whether it’s something hard that we can say, “This is money,” again, “This is dollars and cents”.

-              So let’s start with the easy stuff; what are the tangibles?

-              Well, infrastructure services, for example.

-              You have to make productivity for sales, for example. So yes, there’s intangible benefit, the salesman can go access the information from anywhere. But for you to measure that in dollars, you have to look at the overall productivity, so there is intangibles.

-              If you think about everything you put in the cloud you’re putting it there for some other reason. So let’s pick your SalesForce type application to say, “I don’t go out to choose what’s the ROI of SalesForce.com?” Hopefully you’ve gone through a bit of a process to say, “I need a SalesForce type application, whether it’s CRM, whether it’s SalesForce.com, whether it’s Microsoft Dynamics. Whatever it is, you’re going to go through an analysis to say which one of these meets my business needs, what does the pricing model look like and as part of that conversation, is that hosted on premises or is that in the cloud?

For example, I’ve got this database running on premises. What’s the reason that I’m even looking to move that? What’s my cost currently? What do I hope to achieve if I was to move I to the cloud, then you can start to figure that out. If you just go, “What’s the ROI of the cloud, of moving to the cloud,” I think you can kind of come up with a bunch of numbers but you’ve still got to come back to is that good or bad? Well, compared to what? Is my company going to benefit by moving it?

-              You know, when you look at the client business cases, right, the ones that aren’t built on hard dollar savings are very difficult for organizations to consume right now, right? And I’ve seen this recently, right? We have a large client who’s doing a very large Java cloud type project and you know, all the developer productivity stuff and all the sort of agility, nice, nice, we don’t know our benefits… They’re building a private cloud, mind you, right? But all that stuff, the finance guys, the CFO didn’t want to hear about it, right? He wants to hear about you’re spending 10, 20, 30, X million dollars a year on this today, what’s the new model look like as we go forward, right? And what’s the cost structure of that new model? And the harder the savings, the better the business case, I think, for most clients these day, right? And that’s just, I think an economic climate decision for a lot of people right now.

-              Yeah, by the way, I want to cut in here because I’ve noticed that it depends on what size company or organization or department you’re talking about. Because what I noticed over and over again is that when you’re talking about it in general and the CFO, CIO or whoever’s responsible for the dollars and they’re managing that, they’re just looking at this big cost savings, or not. But they’re looking at that. Like you were saying, it just comes down to… At this level, if I can show a savings, I get a bonus this year, you know? And that’s it, right? But when you look at it at a department level, they almost never look at it that way. They’re almost always saying, “Can I get access to the resource that is going to help me do my job?” they’re almost always looking at it that way. And so that’s why I think we see so much innovation coming up and rarely do we actually see it come down. Is because they’re looking at it from a fundamentally different… They’re more nimble, they’re solving the problem that they need to deal with. And I’ll give you one really obvious example that happened to me.

So I was at PayPal and PayPal operates like a bank mostly and they’re very sensitive about data. And when I was there it was fairly early and they had never used any Software as a Service products, ever. Every piece of information, for the most part, anything related to customers, especially, was internal. The company was growing so fast, though, that our SVP, basically the person second in charge at PayPal, who I reported to, was trying to grow this biz dev team and the sales team to go after big accounts that would use PayPal. Big, giant websites, right, like Wal-Mart. And what was happening was, we put a request for a CRM system, because we were all using Excel, okay? And everyone had their own copy but we had this shared… It was difficult.

And it worked for us when we had like five people but now we had like 20 people. And six months of asking IT, “I want to set up a CRM system,” six months and she’s the SVP. And every month when we’d have our meeting and she’d ask what the status was, her face would get redder and redder and redder. And finally the reason why the IT couldn’t deliver is, quite frankly, PayPal was growing so fast that they were just trying to keep the website up and all that kind of stuff. I mean they just couldn’t deal, so this was like not even on… It wasn’t even registering. So finally, one day, in the meeting, she blew her stack and she just got her laptop right in front of everybody, got her credit card out, SalesForce.com, “I don’t ever want to hear about this problem again.” And it was one of those things where it was like, look, we have a problem that we need to solve, you know? Our corporation is telling us that we can’t do it the way that we need to to be successful because they’re looking at security issues and CIO control and all this kind of stuff. But the departments, they’re the ones who are like kind of oozing out and solving their problems. So when you’re talking about ROI, it’s like are you talking about top down or bottom up? I think it’s a very different story.

-              You have the perfect example. I think what we’re starting to see is customers that have a business need to get something done can literally go to a GoDaddy or to a SalesForce or to an Amazon, punch in their credit card and get stuff up and running. If you’re the CIO, that scares the hell out of people because now you’re going, I’ve got all these rogue projects that I don’t really know anything about that IT either isn’t supporting or can’t support or it’s not important to them. It actually puts IT in a really interesting spot because they become less and less valuable to the business. They just get seen more and more as a hurdle, inhibitor, and people are just going around them and doing this other stuff. So I think you kind of look at it two ways to say, “What are the different projects?” And there does become that element of, it’s going to take us a year to get Siebel or SAP or something deployed in house through this process or I can go on the website and be up in 20 minutes, right?

-              So is one of the ideas that you’re giving there is saying okay, for the non-IT people it gives you a quick go, start, go live?

-              Right, and I think it comes back to a comment you made earlier; it’s a lot harder to quantify that intangible. What’s the cost if it’s up and running tomorrow and we can start using it, versus if we don’t have this for another year?

-              The rapid response is definitely a key item there, right?

-              The question is, are you getting… It depends on where you sit in the organization, right? Because you know, there’s a lot of risk and governance issues that large organizations are going to be scared about, right, frankly, and making those choices. And if you make… You know, there’s a life cycle issue there. You make that decision to go to SalesForce, you’ve made a decision that has ramifications not for just the six people in that little sales team… If you’re a 10,000 person organization and you get 50 different SalesForce deployments with 50 different service providers, you’ve now got a bigger problem, right, that’s been created. And I think that’s the IT Department fear, is everybody goes and just does some coin operated thing and they end up with chaos, right?

-              If you’re running application today internally, you’re going to know roughly what that’s costing you, right? You can figure out what your licenses cost you, you can figure out what your hardware cost you. You might even go into your data center costs or something like that, or an incremental part of your data center costs, including things like power and I don’t know, maybe even water, heating, electricity. You know, there’s all those costs and then there’s the human cost of managing that sort of stuff. And you can go in and calculate those things, right? And I think what you’re really asking is, “Now if I’m going to move into the cloud, what are the equivalent costs?”  And then you’re going to have the migration cost as well, because you’re going to have to tear down the old system and do something with that and that’s going to cost you money. Maybe you already have some costs that you’re not going to get back. And then when you move into the new environment there’s retraining that goes on, people have to learn how to use this new system, perhaps and then you’ve got of course the retooling of the application to get it to run in the cloud. But then once you get all that done, you’re back to just simply calculating what it’s costing to run the application. So you know, at the end of the day I think the hard part is to figure out not the costs of running the system here versus running the system there, it’s all that in between. And then, what is the net effect for your company? Is your company able to operate better, capture more market, run leaner as a result? So you’ve got all these major concepts that you’ve got to go through. Kind of you’re almost evaluating starting a startup.

-              Will you say that for a startup it always makes sense to go cloud?

-              I’ll give you a real quick answer on that. Today I think a lot… I think it does, right? If you’re going to spend startup capital in buying servers and building a data center and running your startup, right, I think that’s not a great business decision.

-              But there is still pay-off at the short term because what if your business is not really viable in two years? You don’t know that yet, you think it is. But you have a two-year, say, window. Now you can invest in fixed investment, old school style, get… Because you may not be big enough. Like if you go with Amazon Web Services, the minimum is still pretty high. Eventually owning the stuff might be better for you but in the short term… It’s like leasing versus buying a car, right? Like (overlapping).

-              Yeah, exactly. I think the conversation ends up being as a startup, what’s your core competence? And so think about it, you’ve got to separate two aspects. One is, what do I need to do just to run a business? So I’m going to say that’s stuff like your e-mail, right? it’s pretty much a given that everyone’s going to need e-mail. So does it make sense for you to own a server, buy licenses, provisionally manage e-mail as a startup? I’m going to say probably not. Unless that’s your core infrastructure. Versus am I building an application? Sure. Do I need infrastructure on premises to do that? Maybe, maybe not. So I think you want to separate out what are the different pieces you’re talking about? Because you have your financials, your CRM system, your e-mail system, your collaboration tools. As a startup I’m not seeing that buying and owning and operating that is necessarily a good investment short (overlapping word).

-              You’ve got to do the spreadsheet, right? There’s no simple answer. You’ve got to do, what’s my volume going to be, what’s it going to cost? Like I’ve talked to a couple of guys here in the city, right, and the CTO, he keeps that spreadsheet. He knows exactly what his per CPU hour is to run his machines and his own little data center relative to what it would cost him. And the reason he wouldn’t move to Amazon at the time was it cost him more at Amazon and he could do it himself.

-              So is it reasonable to say for a small startup, that if they don’t have the ability to measure that type of thing in the first place then it’s viable to go to the cloud just so you can have that measurement because it’s difficult to do that?

-              That infrastructure may still be, just for a startup, out of reach. Infrastructure as a Service. SalesForce is good for anybody because it’s got to that granular point that you can really just one user at a time and that’s good enough.

-              Yeah, if you can find an application, that’s what you need to do in a SAAS environment, it’s almost a no brainer. And by the way, to be really clear, SAAS actually has two parts to it, two types of SAAS. There’s the on demand which usually implies multi-tenancy and then there’s Application Service Provider, ASP model. Which means running one set of code for one customer, okay? And what’s interesting about this is we’re starting to see a resurgence of the ASP model. Because the ASP model was very exciting. Before SAAS had kind of evolved, the ASP model was like this new, cool thing back in like 2000 and it turns out to be not very efficient because you’re basically managing multiple code bases… Well, a different code base for every customer and it didn’t turn out to be very effective. The on demand model where you have one set of code running multiple customers turned out to be much more effective. But, now with Infrastructure as a Service where you get a stored… where you can save a snapshot of an image running on a virtual machine, if you choose to run one of those, that’s essentially an ASP model. And so when you hear people talk about running like WordPress on a cloud, that’s ASP. That’s not a multi-tenant environment.

So we’re starting to see that model come back, which is kind of interesting. So when you’re talking about SAAS, make sure you understand there’s two different models and if you think about it then the advantage of the ASP model is not the same advantage as the Software as a Service model. It’s simply the only advantage of that is you can get it up and running when you want it, like that (snaps fingers), pressing a button. That’s the only advantage, you don’t have the cost advantage. So different needs have different ways of solving them and so I don’t think you… there’s no way you’re going to write one paragraph that says, “Here’s how you calculate ROI for the cloud.” You’re going to have to pick a use case.

-              But I think there are some relevant metrics. I know when I went to school, I went for finance, not comp science, and one of the metrics that was always tossed around but was always abused and profs hated us for it, was the return period. Now for all intents and purposes, the period of return is pretty much irrelevant to a project. That if I’m investing in…

-              That is all variable cost.

-              Exactly. And so if I’m investing in R&D for a new engine, it’s how long after this initial investment are people going to buy enough engines that I will make those funds back? The only time it was really relevant was when there was a hard deadline on that period of return. And I think with the startup idea, you can say… If you’ve got a startup or a small department, if you’ve got VC money or a manager that’s really fast fail, that if it’s not working, it’s done, I think there is a real thing to say, your period of return is going to be really short with Software as a Service. You’ve got to keep paying for it, whereas say I went and bought a $10,000 server , initially I’ve got that huge cost. Eventually, sort of, your increment or your variable cost, or your recurring costs are going to catch up to that server. But it’s that period of can you withstand two, three years of having that initial outlay?

-              Good point. It becomes a question of cap ex versus op ex, right? What are you spending as capital expense up front to get infrastructure, to acquire licenses, versus what’s the operating expense year over year to keep that stuff running? And in a lot of cases on the cloud stuff, you’re removing a lot of that cap ex piece and it all becomes op ex. So one of the interesting pieces that companies look at is, it’s an interesting place to go sell cloud for the guy that has no cap ex, right? if I’ve got no capital money to spend, I can still sell you a solution in the cloud that you just pay monthly revenue for.

-              So it does become an interesting… We can’t expend cap on it but we’ve got operating expense that I can go do whatever I need with.

-              And the variable usage of that, right, that comes, right? So if you’ve got something, a resource that’s going to be used 80 percent, 90 percent of the time, right, then your equation on the operating side of it may change if you have to pay for consumption for that equivalent usage in the cloud, right? And so part of the business case becomes okay, if I’m going to buy this thing and it’s only going to be used five percent of the time, then that’s probably a pretty bad decision to buy it, right? it may be better to rent it and rent it from a cloud or share it, right? You know, drive efficiency through the sharing of the resources.

-              If you talk to enough customers, you’ll run into the customer that says, “I have this business need. I know that doing it myself or buying it is the right way but I’m not going to get budget approval for $1 million of capital expense to go acquire the infrastructure to do it. So I’m going to pay you for the service, even though I know for the long term it’s going to be more expensive.”

-              You’re paying more to reduce risk, though. Because that’s another factor, is risk. Because I mean how many times you hear projects get funded, huge projects and a project doesn’t go that well, the whole thing just gets shit canned after like two years, but they’re out $10 million.

-              That’s associated risk costs? It’s like a risk analysis?

-              Oh, certainly there’s risk… there’s a reward thing, huge, huge. In fact get back to the VC thing, I mean that’s a big, big part of it. VCs love cloud because they don’t have to commit all their money up front. They can say, “I’m going to commit $10 million to you but I’m only going to give you a million now and I’m going to pay that $10 million out over milestones.” But the milestones equate specifically to computing resources based on your success, right? So now they’re like, it’s a win win for them. it’s like hey, I get to only really risk $1 million but I’ve got a reward based on $10 million. So it’s pretty cool.

-              And there’s a flip side of that, because I’ve worked for software startups for a long time and it used to be a software startup would get millions of dollars and all that, and now you’ve never heard that before, but fail cheaply.

-              Fail fast, fail cheaply. And I’m like, whoa, that’s very different. But there’s two reason for that. Basically, you can not put in a lot of money but if you succeed beyond your wildest dreams then the elastic, you know?

-              I’m doing a startup right now and I’m not taking any money.

-              Spend the money on the stuff that matters, right? And that’s your choices here, right? If infrastructure ownership and control is a part of a core complex of  your organization and a critical element of your organization, then you’re going to be more reluctant to move back outside, right? If you win by out computing the competition then you’re going to want to own that infrastructure and you’re going to want to drive it.

-              If you win by having the best SalesForce system then you’re probably going to want to control your SalesForce system, right? But those are decisions, right? If CRM is CRM to me, why would I want to own it and go through all that hassle, right? And those are the decisions that organizations have to make as to what makes sense for them to move to an outsource model.