Thursday, May 17, 2012

Podcast: Kurt Milne discusses how organizations are building clouds


Kurt Milne is the managing director of the IT Process Institute and has been surveying organizations about their cloud adoption. The most common strategy is to leverage existing resources where possible by taking an open, hybrid approach. Kurt is also co-author, along with Andi Mann and Jeanne Morain of Visible Ops: Private Cloud. Kurt discusses:

  • Strategies for cloud adoption
  • How open cloud approaches are proving most popular
  • How organizations are achieving agility through self-service
  • Owning vs. renting capacity
  • How IT decision makers need to look at cloud through a framework of cost, benefit, and risk
  • The need for a portfolio view
Listen to MP3 (0:14:44)
Listen to OGG (0:14:44)

[TRANSCRIPT]



Gordon Haff:  You're listening to the Cloudy Chat Podcast with Gordon Haff.

Gordon:  Hi, this is Gordon Haff, cloud evangelist with Red Hat and I'm out here in Silicon Valley at the Open Cloud Conference. I'm here with Kurt Milne, who is the managing director of the IT Process Institute. Hi, Kurt.

Kurt Milne:  Hello. Glad to be here.

Gordon:  Kurt, could you maybe tell us a little bit about yourself and the IT Process Institute?

Kurt:  The ITPI is an independent research organization. We use empirical evidence‑based studies to try to identify who are the top performing IT organizations and what they do that's different from other folks. We do studies. We write white papers, prescriptive patterns. We also self‑publish the Visible Ops books. A lot of folks have seen the little yellow and black Visible Ops book. We've got Vis Ops Security, and also most recently is Visible Ops Private Cloud.

Gordon:  Kurt, you've started doing some survey work around cloud adoption. Rather than just getting opinions about where things are going and coming up with some numbers, this study that you've been working on really tries to correlate what people are doing with the results they are seeing, which seems pretty interesting. I know it's still pretty early for the analysis of data and finalizing things, but I'm wondering if you could share some of the interesting things that you're finding.

Kurt:  We fielded a survey through the folks at Cloud Camp and also at some other sources. We got responses from about 150 companies that have deployed private or private‑hybrid cloud past the proof‑of‑concept stage. As you mentioned, the goal is to try to look at what were people actually doing. What were some of the pre‑conditions before they started their cloud project? What were some of the key dependencies during their project? Then, what were the results that they achieved? Both positive results and project friction points. The idea is to try to figure out what are those combinations of factors that contribute to cloud project success.

Gordon:  Maybe you could share to start off with, have you found any surprises so far?

Kurt:  Well I think one of the surprises is that there are a lot of organizations, about 40 percent of the respondents, that suggested that their primary cloud strategy or goal is to build an open cloud or a private‑hybrid cloud that leverages their existing assets as much as possible. Another strategic option would be to build what I call a siloed cloud, which would be maybe a cloud that's carved off in the data center for dev and test environment, self‑serve resources for developers, that sort of thing. But the primary strategy was that open cloud, to try to leverage assets as much as possible, and then, tap externally acquired.

Gordon:  That seems to fly a little bit about some of the conventional wisdom at least that we seem to hear out here in Silicon Valley. That the cloud is new and enterprise IT is old, and that the cloud is all about really starting afresh in a greenfield.

Kurt:  Well, I think if you step back and look at it from the CFO or controller's perspective and say, "Why are we renting computing assets from a third party, when we've got underutilized assets internally?" I think that story of, "Let's do what we can with what we've got before we tap external resource pools" ‑‑ I think that still makes business sense.

Gordon:  Great. Yeah. This isn't to say, of course, that people are not making use of Amazon, or they're not putting in new servers, but it really does seem to speak to this idea that we can't just afford to throw everything away and start with a clean sheet of paper. That would be really appealing as an enterprise architect, but it's probably not practical for a lot of enterprises.

Kurt:  Well, there were about 10 percent of the folks that filled out the survey that we got data on that did indicate that that was their cloud strategy, that they're some kind of service provider, that they're building a cloud solution, and that they're not encumbered by existing legacy applications. There are folks out there that have that luxury of being able to start from scratch, and just look at their requirements and not as many constraints, but most organizations have constraints already in place.

Gordon:  Among those enterprises that are building open clouds, open‑hybrid clouds, why are they doing it, and how are they achieving success?

Kurt:  Well, the use cases we looked at that were most frequently deployed by the 40 percent of the companies we looked at that are in that open‑cloud category, the self‑service development and test environments, the self‑service resources ‑‑ so that's really helping achieve an agility state, where folks can tap and get self‑service, on‑demand access to things that maybe they had to go through operations and wait for it previously. There's a self‑service agility element there. But then they're also using the cloud for basic blocking and tackling operations, things like backup, high‑availability disaster recovery as well. Then, we're also seeing a lot of interest, based on the survey, in scale‑up and scale‑out capabilities that may be difficult in a more traditional, static IT environment.

Gordon:  It's interesting, although it's very consistent with a lot of other data we've seen, that cost savings does not seem to be the primary driver here. Not many a CIO is going to tell you, "Oh, I can spend as much money as I want," but that doesn't really seem to be the driving force behind people adopting cloud.

Kurt:  I agree with you. I think there are some efficiency gains, when we look at what were the results of your cloud effort ‑‑ more development efficiency, more run‑time or operations efficiency. I think there are improvements there, but it doesn't really suggest that companies are tapping Amazon and others just for the cost‑savings aspects. It's really the process efficiencies in combination with those agility factors that I think is driving the adoption.

Gordon:  Let's talk about hybrid clouds. I think sometimes "hybrid" has been taken to mean this auto‑magical, super‑fast, dynamic shifting of workloads among clouds, which, frankly, I haven't seen happening very much. But I do see, still, a lot of interest in being able to move between clouds, even if it's done at an administrative level as opposed to an automatic, load‑balancing way.

Kurt:  Yeah. Interestingly, I think in the last couple of years, as cloud has gotten a lot of attention, people have talked about this bursting concept, where you're going to have a workload internally. Then if you have an unexpected or even a planned usage spike that you'll be tapping external resources, but that was the lowest prevalence. When we asked what folks were using their cloud for, that had the lowest percentage of companies indicating that that's what they were using it for. Whereas starting in the cloud as a prototype environment, doing dev and test work and then, once the workload stabilized, bringing it back in‑house, had a much higher rate of response versus bursting out.

Gordon:  That seems to be consistent with particularly what some companies, like Zynga, for example, are doing, this idea that you own the base and you rent the peak.

Kurt:  I agree. I think that is the Zynga model. I think it becomes an owning versus a renting kind of decision, right? In some cases, it makes sense to rent resources, and then, in other cases, it makes sense to purchase it and utilize the resources. I think, in Zynga's case and a lot of the respondents in the survey, there are times when you rent resources, but if it becomes stable and predictable, then it makes sense to buy the assets needed to support those over time.

Gordon:  The interesting point you're making is what you're saying is this becomes, essentially, a financial or a capital‑budgeting decision at that point, OPEX versus CAPEX. When a company decides to rent or lease or buy company cars, for example, they don't have to get a different kind of car, depending upon what their financial model looks like, and that hasn't always been the case in the IT industry.

Kurt:  It's interesting you mention that because I know with Zynga, one of the driving factors of their decision to build a private cloud was that they wanted some capital assets that they could depreciate. I think there's always this talk in IT of operationalizing costs, converting from a capital‑asset cost to an ongoing operations cost. In their case, with no assets, a completely virtual company, if you will, they were actually looking at getting assets to depreciate as one of the drivers for building something, which I thought was opposite of what a lot of people talk about.

Gordon:  The other thing you see in this idea of pay‑per‑use and all. People and companies are very big on pay‑per‑use, as long as it means they pay less than they would have otherwise. They're not so big on pay‑per‑use if it means paying more.

Kurt:  I think the other aspect to that is I think companies want to pay‑per‑use if that cost is tied to revenue. If you've got a model where, adding an incremental customer, you have an incremental cost associated with delivering service for that customer, I think the financial folks like that kind of arrangement. In some cases, they're willing to pay more for service, if that service is tied to revenue that's scaling up and down versus maybe paying less for a fixed asset, where that cost isn't tied to revenue. I think that goes into the decision, as well.

Gordon:  That's also a fairly traditional part of corporate finance. You're willing, at some level, to pay at least a small premium for being able to match up your revenue streams and your costs streams. Basically, this isn't any different from that. Any final points you'd like to make?

Kurt:  Well, I think it's interesting. We're entering the podcast, here, talking about more traditional management methods, but I really do see a lot of these cloud concepts enable new dynamic IT capabilities. IT decision makers continue to be very pragmatic and need to look at all of these things in a framework of benefit, cost and risk. I think these new capabilities have different cost and risk factors, but there's no magic here. It all needs to be viewed from a pragmatic lens, in order to make decisions on what's the best path forward.

Gordon:  That's a great point. I think it's really exciting, a lot of this stuff that's happening in cloud, new capabilities, all of this API‑based, modular computing, self‑organizing, what have you. I think it's all great, mind you. I'm certainly not suggesting otherwise. But we do, at some level, need to balance all the new application style enthusiasm, by the fact that, in most cases, we have run the business applications that we probably don't want to just abandon or say they're old legacy apps that we really can't do anything with and not concentrate on them any longer. Those are the apps that are running the business, and, in many cases, there are things that open‑hybrid cloud management can do to make them more flexible, as well.

Kurt:  Yeah, I think it becomes a portfolio‑view approach that organizations need to look at their whole suite of applications, determine what the best fit is from an environmental standpoint. They can't throw out legacy apps. The apps are being used by folks in the business to do business functions. I think creating some framework of being able to evaluate what's the business objective? What are the constraints? What are the architectural options that make sense across a whole range of physical, virtual and cloud environments? I think it's going to be a mixed model. I don't think any one of those is going to prevail or any one of those totally go away.

Gordon:  You go into quite a bit of methodology associated with that in your last book.

Kurt:  That's right. The plug for the book, the "Visible Ops Private Cloud: From Virtualization to Private Cloud in Four Practical Steps," was based on interviews of about 30 IT organizations that had deployed some form of private cloud. We were really trying to capture some of the lessons learned. What do you know now that you wish you had known at the beginning of the project? What were really the key success factors? Then, put those into a step‑wise methodology that we think any organization should at least consider some of the factors there as they develop their cloud strategy.

Gordon:  Great. Well, thanks very much, Kurt. Been good talking to you.

Kurt:  All right. Thanks, Gordon. I enjoyed it.

Gordon:  Bye everyone.

No comments: