This is a transcription of an interview between James Urquhart, Market Strategist and Technology Evangelist for Cloud Computing and Visualized Data Centers at Cisco Systems & author of the CNET blog, “The Wisdom of Clouds”, with Wikibon Co-Founder David Vellante and SiliconAngle.Com Founder John Furrier, conducted at the Strata Putting Data to Work Conference February 1-3, 2011, in Santa Clara, Calif. The interview was originally broadcast live on SiliconAngle.TV and is available as a recording. James Urquhart was named one of the 10 most influential people in cloud computing by MIT Technology Review July/August 2009.
JF: We're back inside the Cube with fellow cloud clubber James Urquhart who's a friend, distinguished writer for Cnet, and also works for Cisco Systems, knows a bit about clouds and been in the trenches about the whole cloud phenomenon and technology conversion networking at Cisco. [He] has followed the pace of change and hype on cloud from Amazon, public, to full-on enterprise adoption. Now big data is kind of stacked on top of cloud and everybody's seen the light, it's a mobile world. Cloud's the real deal. No more big debates, public/private, okay, what do you do?
JU: I think the ultimate thing it comes down to is more and more what's happening is: Infrastructure-as-a-service was a big “Yeah! This is awesome!” What we have now is a little more, “Wow, the services world that is possible in the cloud is really broad.”
So I wrote some posts recently saying I think the household names in cloud long term aren't going to be the folks like Amazon!, although they have time to change their business model as they go. But it's going to be folks like Microsoft and Google and others who will end up having that suite of IT service capability. Say I'm a new business – I'm opening a new business in a town – and I want some IT to back that up. I go to one of these sites, I set up an account, I check off that I need the retail system, I need the finance system, and all of that integration is working for you, all of that ability to process the data as it comes up and is generated is readily available for you as a service capability. One of the things that Cloud as a model does is change that from how do we get people to pay a yearly fee or license fee to get that kind of capability and changes it to “Now I can let people try it & play with stuff with it. I can even offer it free for a period of time.” It really opens up the business model capability.
The other thing it lets you do is try new things and fail, which is different from how IT has been. In the past you really had to justify the upfront cost to give something a try. Now what we're seeing, and I think this is something that we're seeing on the iPhone that is very exciting, is developers going out and cranking something that they hope 10 million people with sign up for. And you've got a sort of lottery system underway with it, but those that succeed succeed big, and the ones that don't … It wasn't that expensive to fail.
JF: Where's your head at these days in terms of levels of excitement. You've always been focused in on the trends and monitoring and evaluating it, but where's your head at right now. It's a good environment, we're seeing people literally getting checks from VCs and angels – here's $800,000, go play – but the enterprise is pretty hot. Do you feel good about this? Cautious?
JU: I think the economics are just too compelling in the long term for the cloud model not to play a major if not central role. And what I mean by “cloud” in that statement is there will be a hybrid cloud model ultimately, but a large percentage of the compute services and the basic storage services will probably be in the public cloud. The reasons for that are really clear when you look at the work that folks like Joe Wyman has done on the cloudeconomics.com site. You look at the math and it's really about the utility costs & how cheaply can the provider provide it to you vs you providing it for yourself. And man, the numbers...It doesn't take long until you start seeing the numbers really shift if you have a large-scale provider.
The problem is today when we start talking about cloud provider, with a few exceptions in the SaaS space we're really thinking in terms of infrastructure and platform most of the time. I think those are critical pieces of the story, but I think the long term story really starts to become – and I always forget to mention folks like Zoho and others that are smaller that are in that space – but this ability to provide some sort of integrated capability around business. And on the data processing side a huge other market that drives cloud capability and the consumption of resources. But ultimately the service that will be presented won't be about servers or storage, it will be about BI, data filtering, mashups, and these kinds of things.
JF: So ??? is a big topic obviously. We rent space from Cloudera and were at Hadoop World. But you also have been watching deeply in virtualization. And virtualization, these kinds of things are under the covers, powering all this utility. And the theme here is value, business value, a lot of discussion around that. What proof-points are you seeing that the business value is materializing in this cloud/mobile/data world?
JU: The easiest one is to look at the volume of business on folks like Amazon and Rack Space amd others & how quickly that's growing. Right now most of the tools are build right on the virtual machine and on up, especially in analytics & data gathering & other things. There's not a lot of pre-built services. There are some Hadoop capabilities that are offered by some of the providers. But ultimately setting that up and putting the code in to do the processing, that's is a lot of work on your side.
So part of what I think you're seeing is that because these are bursty – they use a lot of capacity for a period of time to do some analytics & then you shut that down – what's really interesting is the amount of volume actually isn't reflective of the cumulative growth, but it's a sign that there's a continued steady consumption of these resources for these purposes. While Web sites and other applications do play a fairly major role in Infrastructure-as-a-Service, data processing for me is maybe the top winning, killer app.
DV: So the angle you're taking, normally we hear from a lot of the traditional enterprise companies is that the public cloud – it's a service-level issue. Service levels are – Joe Tucci recently described it as “We'll do our best, but if we fail don't call us” kind of thing. That's the angle that some companies are taking, but you're saying it's more than that, it's about the application, the solution, and that's what's going to drive cloud adoption.
JU: Right now we're in this world where developers are discovering cloud. Therefore it needs to be at a level where developers can consume it, and that's the thing that's driving it. If you go to enterprises and ask “Why are you building a private cloud” or a cloud at all, it's really to provide a better capability to their internal developers as much as anything. There is some migration of workloads into cloud models, but the economics of making a change to a stable workload that's running in your data center just to get it into a cloud, if it doesn't have this kind of dynamic just doesn't make sense. It's really not all that horribly compelling. But what is compelling is that there are companies out there with dev/text laps that take up two floors of an office building where most of the servers are sitting there waiting for the next release and next build of the software before they get consumed. That is the kind of thing where they say, “If we could actually consolidate those resources and make it self-serve to consume them and maybe even be able to burst out to public cloud environments, there's a lot of value there.”
I think the security question is still legitimate, and I think private clouds will be built because there are still some things you can't take to public clouds legally or there are technical constraints around security or other elements.
DV: Or maybe culturally.
JU: Or maybe culturally or whatever it is. And I think it's perfectly legitimate to go in that direction. I think, though, that when you look at the economics of the situation, the position that we take & that I've taken on the blog is to say you can't really look at individual development successes & say, “There it is, everybody's going to move to that model.” What you have to understand is when do the economics make sense to change your model and what are the advantages that the new model brings in. When you make that statement, startups instantly make sense, small businesses it largely right away makes sense. When you get to enterprises that have a sunk cost in an existing infrastructure, now you get yourself into a situation where it is really difficult to justify making a move just to get on the new model.
DV: Yes, so then the question is what is the best path for those guys. CIOs I talked to say, “Look, I don't care what you call it – private cloud, public cloud, IT-as-a-service – it doesn't really matter to me. But I do want” what they describe I would call as cloud-like attributes. They want pay as you go, they don't want to pay a big up-front capital expense, they want elasticity, they want all those things, that flexibility. So the question is can they get that from a so-called private cloud? Do they really have to, or does the industry just have to close that gap? What's your take on that? What makes the private cloud a cloud?
JU: I think what makes the private cloud a cloud is eliminating bureaucracy of the IT organization having to make capacity decisions and having to make operational decisions around the allocation of resources. I think it's that self-service capability; some form of ability to indicate back to the business what the cost of the cost of consumption of resources is. Ideally I'd love to see charge-back and show-back, but there are issues with that that have been pointed out by some bloggers recently. And then there is that democratization where you say, “Look, right now if you want some servers to be put in the data center you have to go through a committee, you have to go all the way up to the CFO's and get his approval.” If we could just turn around and say, “Look, go on a Web site and make your request, and ideally within 5 minutes we send a pointer back to your server.” Maybe in some businesses because there are security issues or resource provisioning or even cost provisioning issues you might take a little bit longer because for some process. But ideally you are still reducing it from days to minutes or days to hours, which is still a huge benefit. There's been some recent work that shows that the faster you can get the resources in the hands of those who want it the more money you save overall. So that's a big part again of the story. If you can get bureaucracy out of the IT organization in terms of provisioning and then in terms of operational decisions moving forward but still provide the services around backup and all those things, you're going to come out way ahead.
DV: Okay so you can check the agility box on that one, and there's clearly a cost implication in there. How about the whole capex vs. opex thing? Is that something that organizations have to resolve? Do they have to go to service providers? Do they just become the service provider?
JU: I think that's the ultimate thing. The capex vs opex story is is one in which at some point in time you're paying for the equipment anyway. At some point in time you're paying for somebody who bought a server, and they want to make a profit off it. It's a cash flow story, it's not a savings story. In fact....
DV: It might even be more expensive.
JU: One of the things that happens is the so-called Devins paradox, where in fact the cheaper you make the resource the more resource ends up getting consumed anyway, and you end up spending more overall on that capability than you would have when it was more expensive. So from that perspective you can't look at as a pure cost savings story. Then if you are going to make that capital expense up front, and do the private sourcing, you have to understand what the justification and the economics is. For me what it comes down to is the more applications you can throw into a cloud the better the economics are. If you have a silo data center where your utilization rates are bad or the only way you are fixing it is by statically provisioning vertical machines to do fractional ownership of the physical server, but you don't have much flexibility, then you would still have huge costs in operations that don't really need to be there. If you can get automation into the data center – and cloud is an operations model, it is not a technology, it is a way to apply technology to problems. So from that perspective if you can change your operational model to take advantage of the economics of shared resource pools and shared capacity, you may not be as cheap as Amazon in what you're doing, but you will run a lot cheaper than you have been in the past, and you will be able to maintain, certainly, security, compliance parameters, certain cultural parameters.
DV: So it's an asset utilization issue. As long as you've got that meaningful improvement in asset utilization, you may, in fact, even be better off with a capex rather than an opex model. Maybe.
JU: Again, if you get really good at operating those systems that you paid for and ultimately either the business is paying you back and/or you can show the savings to the business in the agility and all the other things we talked about earlier, then all of a sudden it just doesn't look as silly any more.
DV: What about enterprise meets cloud meets big data? We saw...you remember the Berkeley paper “Above the Clouds” that came out and said, “Inside of a data center, that's not cloud”? All the enterprise companies that I work with said, “Woe,woe,woe, wait a minute. We're cloud, dude, okay?” and the blogs came out. That's fine, it's good marketing. We're kind of seeing the same thing with big data now. What's your angle on the enterprise meets big data meets cloud?
JU: To me, personally, it comes down to a lot of the same economic arguments where you say, ultimately what is your sunk cost today in terms of the infrastructure you have today? How much does it cost you to change the processes, the data extraction, the aggregation, algorithms that you have in place around the data, or add the new ones you want to add to it? What does it cost you in terms of the infrastructure move that you have already sunk into infrastructure and support that you're now going to have to sell or get rid of in some way as you move to an external environment?
Frankly, though, I think there's big classes of applications when you talk about big data that were unbelievably expensive for an enterprise to take on that now are not nearly the big deal. An example I like to give is I told the story early on in my cloud writing about Eli Lilly probably two or three years ago now.
DV: A pretty famous example.
JU: A pretty famous example. Basically what they said was now “We set up some accounts with Amazon and told our researchers here, we set all this up so you can go and try and play with this, and it looks like it could be really useful to you.” They didn't touch it. So they went on a road show and showed each of the researchers one-by-one what is it that you need to do to use it, & they started to play with it. And about 3 or 4 months later the director of the research group or VP or whatever he was came back to them and said, “You know what? Thanks to this we're seriously reconsidering the way that we acquire new drugs & do research on new drugs at Eli Lilly.: Now instead of having to look for a start-up that's taken all the risk up-front already and found some drug for us to take over and finish the trials and get it out there, we have all this data from our previous clinical trials that we can go hunting through for a certain hypothesis and certain things to solve the problem. And by the way if we run something for a couple of month on a few hundred servers & it doesn't work out, we're a few 10s of thousands of dollars maybe. As opposed to before we had to justify all that equipment up front from the clinical trial budget. So it costs us $6 million to do that processing, & there's no way we would ever get it approved.
DV: Yes, to your experimentation piece and risk reduction and business value.
JU: I'm extremely exciting about the cost reduction that the cloud brings on. And I think that for big data, we have all this stuff, and the processing's not that expensive. But the execution of the processing is incredibly expensive if we have to own the infrastructure. Now we can go out there and just consume Hadoop service with a bunch of instances underneath it, and just get that out of the way. I think that's one of the exciting things about cloud.
I'm really interested in conferences like this one and other things we're talking about. I know very little about the consumer space. I'm excited about what I am going to be able to know about myself or others around me that I never knew before.
DV: We just had a really interesting...were you here last night & did you see the guys who won the startup award? The Bill Guard guys? Talk about a great consumer app, basically they can if you put your credit card statement data in there they'll match it up with other credit cards to find fraudulent charges. So you don't have to look through your card statements any more, which half the time you don't do anyway. Talk about a great consumer application.
So how about the implications this is having on the industry structure? It kind of used to be that you had Cisco doing the networking thing, EMC doing the storage thing, and Oracle doing the database thing, and IBM and HP maybe doing their things. Now it is a whole world coming together. You have 4-5 companies driving the trends in the enterprise. And then over here, in what I like to call sometimes the “true cloud”, no pejorative intended, but I think you know what I mean by that, is you've got a startup explosion. And those two worlds seem to be coming together, whether it's through acquisition or maybe even some kind of organic growth. What are you seeing there in terms of just the industry structure & consolidation?
JU: I think as a member of a systems company I will be brutally honest and say I think that there are aspects of this that have taken a little while to learn from a business perspective and say wow, there's a real effect there. I came into Cisco knowing there was an effect and came in to help drive some understanding of that. But I think what you're seeing in fact are some really innovative ways to attack several problems. The first is somebody's got to build the infrastructure that's underneath these things.
DV: The plumbing.
JU: Some of the real cloud guys have gone out there & done their own thing, & put together their own pieces and parts, & put together their own software & done all of that. Others are very interested in competing in the space and not having to have that skill set in order to do that. So there's startups that I think are really exciting and driving what you can do with infrastructure. But I think the systems companies have gotten very smart about understanding where there's value to be added into the picture from an infrastructure perspective & how to package things together to make them more consumable in a cloud scale model. So that's the first thing.
DV: That's the homogeneity aspect of it, right?
JU: Yes, one of the things is as much as we hate to go back to, you know, a single sourcing model. I think the truth is cloud drives heterogeneous workloads on a homogeneous infrastructure. That's the model that works really well, because the more you can pool your resources across multiple services even, the more effective your cloud is going to be.
The thing I think is really cool is to watch how the startups have effected the industry in terms of the perception of what are the valid services to offer online or even within a business in support of cloud consumption. So to me the really exciting space right now is first of all an area that never had much support which is application operation. So if you're using cloud you aren't operating the service or the infrastructure, you're operating the application within that cloud, or multiple clouds. So how do you do that? So folks like Rightscale & Stratus – those guys are showing the way on a whole big market space that I think is beautiful. I look at that and I go, “Man if I was doing my development 10 years ago, love to have a service like that now that I look back.”
But the other side of the space is I think the next big growth opportunity is those who want to build services to offer to others. Right now they build them on an Amazon and consume Amazon services. But for those who already have an infrastructure or those who would like a model where they would like to distribute their service across multiple cloud providers, so they can either do that for best price or so they can be in multiple locations and have multiple markets to go after, I think there's a huge opportunity for simplifying the world of service development in the cloud without being tied to the infrastructure – decoupling the infrastructure from the services.
So that decoupling is one of the opportunities moving forward that I think the systems companies have to participate in, because if they maintain that lock they end up saying, “Our definition of cloud is our stuff.” I don't think that's good for anybody.
DV: That is the definition of some companies' cloud.
JU: I think it is.
DV: And other companies are saying, “Our definition of cloud is our stuff combined with their stuff and their stuff and their stuff & these guys are our friends and those guys aren't.” So what's really the difference?
JU: Well I think the part of it though is – I don't want to sit here and hype Cisco. I think what we are looking at is difference between saying there's an ecosystem out there, & that ecosystem for you to be able to consumer it, the whole dream of the market in the cloud space for all the different kinds of services that you can choose from on the fly. It's really difficult to get to when we say, “When we say services we mean these data centers over here.”
One of the things Cisco's been very careful about is to say, “We aren't offering infrastructure services ourselves for a really strong reason, which is because we don't want to compete with the service provider people that we know we can help be incredibly successful.”
DV: We had Howard Elias on two weeks ago, we were talking to him in New York City. He runs EMC's services group. He said the same thing. Joe Tucci actually said the words, “On this one, I'm going to speak for John Chambers. We will not compete with our service providers.” Now they obviously have, and I presume Cisco does too, some little experiments there and there. But ….
JU: Yes, we have some services, but where we have services we have to look at those and legitimately ask ourselves the question, “Is this the business we should be in, or is this the business that we should be helping our service providers be in.” I think you'll actually see us compete with ourselves even in the few services that we already do, where we're offering software to help people be in that space.
But what it boils down to to me is ultimately if we don't open the ecosystem up a little bit what's going to happen is these guys who are building open source software on the cheapest X86 hardware and the cheapest switches they can find on the market & doing a lot of engineering around trying to glue those things together and make them work as as whole, and then the ongoing operations around maintaining that and keeping up with the technologies & so on. Those guys are actually going to start chipping away 'cause what's going to happen is ultimately you're going to have an ecosystem develop somewhere. And if that ecosystem is open stack on cheap, cheap hardware without any support from the service systems companies at all, it's going to be a much more open feeling environment to developers, to enterprises, much less of a lock-in feel thing. And I think it's really difficult to find yourself in a position where you're not prepared to support that.
DV: So you've got to be prepared to change your business model essentially to make the service providers a key partner, make sure your selling costs go down concomitant with....
JU: Or that we offer a value that they can turn around and get buy-in from their customers as well.
DV: Right, because that is a potentially disruptive model but not as disruptive as those open source guys coming in. So we're here with James Urquhart. So James, how do people get to your blog?
JU: The blog is on Cnet: news.cnet.com/the-wisdom-of-clouds. You can also find me on the blogs.cnet.com page right at the bottom.