Interview conducted by Vinod Baya
Erich Clementi was named general manager of Enterprise Initiatives at IBM in October 2008. Prior to that, he was general manager of the Business Systems division in IBM’s Systems and Technology Group, where he was responsible for providing small and medium-sized businesses with systems and storage solutions—solutions that later became some of the first pieces of IBM’s cloud computing portfolio, called Smart Business. Since the 1990s, Clementi has worked in several other roles at IBM, including general manager of IBM’s Managed Business Process Services division, general manager of IBM’s System z division, and leadership roles in corporate strategy, marketing, and sales.
Dr. Irving Wladawsky-Berger is chairman emeritus of the IBM Academy of Technology and visiting professor of Engineering Systems at the Massachusetts Institute of Technology (MIT). Prior to retiring from IBM in June 2007, Wladawsky-Berger was responsible for identifying emerging technologies and marketplace developments critical to the future of the IT industry, and organizing appropriate activities in and outside IBM to capitalize on them. In his emeritus role with the IBM Academy, he participates in a number of technical strategy and innovation initiatives. At MIT, he is involved in multidisciplinary research and teaching activities focused on how information technologies are helping transform business organizations and the institutions of society.
In this interview, Clementi and Wladawsky-Berger share their insights on how cloud computing borrows from the principles of the mainframe era, but in doing so, it brings greater flexibility to the IT infrastructure and applications. They also forecast how, with standardization and mass customization, clould computing is changing the interface between the service providers and service consumers.
PwC: How much disruption does cloud computing represent when compared with the Internet?
IWB: I think cloud is the name we are giving to a new computing model in IT. That’s important, because if you look at the last 50 to 60 years of the IT industry, we have had only two previous computing models. We’ve had the central model, where the mainframe was the biggest product. We’ve had the client-server model, where the Windows- and Intel-based computer was the biggest product. And now cloud is emerging as a new computing model.
The central model was based on mainframes and supercomputing built for sharing the machines with multiple applications and users. Acquisition was very much controlled by a central group, since they were spending millions of dollars and were supporting lots of users. The applications were back-office, mission-critical applications. The central group didn’t let many applications onto the machine, since one application could bring down everything. They had very strict guidelines about what you’re allowed to have in the machine. That was the mainframe model.
In the client-server model, instead of the systems being optimized for sharing, now they were optimized for low cost, because they were built out of PC and workstation technologies. Instead of a central organization, departments did their own acquisitions, which they found appealing because they didn’t have to go to a central bureaucracy. Since these systems were built out of microprocessors, you tended to run a dedicated application on them. As a result of that, there was a lot more flexibility in what applications you could run. So there was a lot of innovation in new applications, because it was so much easier to get the applications in.
Now, a different model is emerging. The cloud model is an Internet-based model. It has taken a while to work itself out, but we realized that something started to change even in the late ’90s. With the Internet, you had so many more users that all of a sudden the servers got far bigger, people started putting them into central computing data centers, and they had to pay a lot more attention to scalability, systems management, and things like that. It started to change with the Internet.
EC: IBM believes this is a fundamental shift in the way services are consumed and delivered. We have seen computing model shifts before, but they didn’t change the IT industry’s own business model the way cloud computing is. We sold a centralized solution in the “mainframe era,” and then we all sold a decentralized solution when the client-server model became mainstream. This move here is different, because it changes the business model on how IT value is consumed. And that is a very significant shift.
It has similarities in impact to the Internet in terms of deployment. Everybody loved the Internet for content, but there were certainly concerns over security, privacy, control, and processes when it first came on the scene. The CIO was nervous. Similar sentiments are true today with cloud.
PwC: It is often said that many of the characteristics of cloud computing have been with us from the mainframe days. What is new now that cloud computing is gaining traction and interest?
EC: Coupled with the delivery, I’d say the first new element of cloud computing is the self-service nature of it. Think about what ATMs [automated teller machines] did to the finance industry. Initially, ATMs were nothing other than cash dispensing machines, and you could say that they just substituted the human interface with a machine interface. But that is a superficial way of looking at it because the human interface did much more. The teller was a built-in security system because the teller identified you and handed you the money because he or she knew you. The teller also managed the whole process of restocking, of accounting. By substituting that human for a machine, it forced the standardization of the entire process within the organization. Then when customers wanted to get their money from any bank ATM and not just their own bank, messaging and the transactions had to be standardized across the entire industry. Otherwise it couldn’t work. So this change standardized the process around cash.
Cloud computing applies the same discipline to IT and changes the way you consume. So far, IT has been something like an art. Cloud computing is doing to the IT supply chain what Henry Ford did with the conveyor belt. Otherwise you cannot explain the economics.
The second new aspect of cloud computing is a different business model. I hesitate to go to variable pricing directly because there are different levels of variability. Many enterprise customers will probably want to have a subscription or some fixed, guaranteed fee and then some variability—but a higher variability for flexibility than before. The third element that is coming into play is the hybrid nature of the sourcing. Until recently, you basically had to have all the technology in your department to consume IT. This is what created the IT industry. Now there will be an increasing number of services that you can consume without actually being in charge of delivering them.
IWB: The rise of mobile computing and sensor-based computing makes this new. By mobile computing I mean BlackBerry and iPhone devices and e-book readers with Amazon.com and others, and now netbooks.
At some level, one of the most tangible differences between central computing and client-server computing was going from text-based terminals used by a relatively small number of people to PCs with GUIs [graphical user interfaces] used by many, many more people. Now these new mobile devices, let alone all the sensor stuff, is orders of magnitude more people, and more things accessing the services.
Also, in the previous two eras, the services were relatively custom. With cloud, the need to support billions of people has put us into the mass customization era, which, by the way, the telcos [telephone companies] know how to do very well. Mass customization is sort of a new concept in the IT world. It’s not a new concept to the telcos or cable. In fact, I would even say retail companies such as Amazon.com and Netflix are all about mass customization. Notice everything Amazon.com does is mass customized, and it’s self-provisioned, but it’s self-provisioned within the boundaries that they give you.
I also think that you need cloud now because the client things that you’re supporting are now exploding, if you include sensors, into the trillions. The central mainframe model will be way too expensive for this, and in the client-server model, the system management costs would be too high. So cloud had to be invented to handle this explosion of new things that come into it.
PwC: What existing problem is cloud computing solving that enterprises should pay attention to?
EC: What is ultimately driving cloud computing is a crisis of complexity, driven by the distributed computing model. So why did the distributed computing models supplant to some extent or complement to a big extent the centralized one? The centralized computing model was too rigid. But before enterprises realized it, they had server sprawl, inefficiency, and low utilization, and soon they lost the flexibility of the distributed model. Remember, there wasn’t a service problem when it was centralized. There was no inefficiency, because it was highly utilized. But the perceived cheapness of the distributed environment came back to haunt us.
So cloud computing now offers an opportunity to deal with this complexity, and guess how cloud computing is doing it? It is doing it with the same virtues, if you want, that the centralized computing had, but on top, giving flexibility. The virtues are the virtualization, the standardization, the automation, the discipline. So in that sense, from an infrastructure point of view, for sure, cloud computing modernizes the legacy.
IWB: Reduction in complexity is a huge part of cloud computing. This is a very important point, because what cloud is using is identical to what the telcos have done for their services, which is standardization and mass customization. In other words, if you say, “There is a general-purpose computer underneath. I can do whatever I want to do. Why do you tell me I can have only an 18-minute plan or a 120-minute plan? I want a 93-minute plan.” It is because the cost of administering a plan that’s different from the available options is so high that it’s not worth giving it to you. It would be overwhelming and add to the complexity of their operations. They do their segmentation, they give you these plans, and if you don’t like it, you’re into custom territory but you pay a lot more.
PwC: What will be the impact of the current economic environment on cloud computing, if any?
EC: One indication on how to read this is that the enabling technologies for cloud computing have been in the making for a very long time—virtualization, service orientation, automation, the networking, the bandwidth characteristics, the whole infrastructure. This has been in development for 40 years. People have been talking about putting together something that virtually works like a mainframe for a long time. So this is something that was inevitable in its economic impact. Now you come to a moment in which cost focus is heightened due to the economy, so this will accelerate the demand for these efficiencies and maybe it will force some second look at what level of customization you really need.
IWB: One of the things that makes new models happen is cultural resistance. People will say, “I’ve always done it this way. Why are you asking me to change?” Now, if I say, “Well, because it’s much cheaper,” a customer might say, “Well, but— it’s not as good as what I did before” or “It won’t work for us.” But when you have a crisis, all of a sudden those excuses don’t count anymore. If you can really get something done that’s good enough and cheaper, then people will pay a lot more attention to it. I think the tough economic situation will push companies to be much more disciplined about making these changes.
PwC: How will the adoption of or migration to cloud-based services take place?
EC: First, computing models impose themselves for economic reasons. If the economics are right, it’s going to be successful. If it’s not, it’s not. Second, things take time. When you analyze the cost of delivery, you need to do it for a specific service or workload. You will discover that the knee in the cost curve occurs at different points for different services, so that is why in the industry everybody says the SMB—the small and medium-sized business that can never get to that knee in the curve—is the absolute first that will go for services delivered through cloud computing. They also happen to be the users that, while needing a full range of services, usually have fewer requirements for customization or configuration.
There is a whole set of services where you can get very, very significant economic benefit from private clouds. Workloads that might be regulated, have security requirements beyond what is available, or that need to undergo significant migration costs won’t work well for a public cloud model. But you might be able to reap significant economics in a private cloud. For other workloads that have massive economies of scale, you might go directly outside. Take high-performance computing, for example. High-performance computing has a characteristic that when you really need it, you never have enough power, and most of the time you have an issue with the economics of it, because you simply cannot load the machinery sufficiently. So what better model than going to a public or hybrid cloud? Where the real cost benefits are, where the cost elements come to the asymptotic range, is very different for different workloads. That is our experience.
IWB: I think client-server may have been an interim solution for the more front-end kinds of services. It may be interim because in many ways, cloud is a kind of scaled-out mainframe. If you look at classic mainframes as being very good for scale-up, SMP [symmetric multiprocessing] kinds of applications, what clouds have done is let you consolidate all these highly distributed applications in a more centrally managed, very disciplined environment, but using commodity technology. Since it’s scale-out, you don’t need to pay the kind of sharing costs that you have to pay with SMPs, and you get much higher utilization of your hardware, much lower system management costs, and far better scalability. So you get things that people associate with mainframes, with central computing, but now applied to this commodity-based scale-up model.
PwC: All enterprises have a significant legacy base of applications. How should enterprises look at the cloud computing opportunity with reference to the base of solutions in which they’ve already invested?
EC: Take your workloads and put them into three buckets. First are the workloads that for reasons of security, migration costs, maybe regulation, or risk are not the primary target for cloud computing. Second, there are workloads that could benefit both from the technology model and from the potential outsourced delivery model. Take those first. For example, 30 to 50 percent of servers are used for testing development, and these systems notoriously have single-digit utilization rates. You can move that whole workload, which usually has fewer security and viability requirements—they’re not so stringent—and reap a host of benefits. Similarly, collaboration workloads can benefit from this kind of scale.
Then there is a third set of workloads that you are not doing today, simply because you never thought you could afford it. But with cloud, these now become possible. For example, take the risk analysis done in financial services. Today it is largely done on a stochastic basis. Why? For two reasons. Either the customer would have to buy such an amount of computing that it wouldn’t be economical, or it would take so long that they would realize the risk after the event happened.
But if the customer could get sufficient computing power to do the analysis in time and at a decent price, they would most probably do it. Think about inline analytics. Think about risk management. Think about extreme high-performance computing. None of these usually have stringent requirements in terms of security, compliance, or migration costs. Remember, when I say stringent, regulation, and security, all of this is a statement in time.
PwC: Customer lock-in—whether it’s an application or data or other elements of IT—has surfaced as one of the big concerns with clouds. Customer lock-in has also plagued users for a long time, with traditional delivery models. Does cloud represent another round of lock-in of a different nature or a freedom from lock-in?
EC: We cannot allow lock-in for cloud computing. I think we need two things. We need interoperability and we need standards. Every maturing industry evolves around standards. The lock-in in the past was in part driven by the fact that there was nothing else. In a nascent industry, that’s one thing. Lock-in could be an inhibitor to its growth.
Customers would want to know how they can get the data out of the cloud and how they would interoperate. The industry needs to know what they find on the other clouds—for example, they can communicate with a cloud, exchange data, collaborate with a cloud, and there are mechanisms to do these. For sure we are going to push for openness and interoperability.
IWB: We will prevent lock-in if we agree on standards. With the Web, there is no lock-in, because everybody agreed to a common set of standards, such as HTTP [HyperText Transport Protocol] and others. With cloud, this is the perfect time to come together around standards. We must have standards, and we may not even know all the places we need standards. That’s why we need the industries to work together and identify and start developing them.
PwC: If standards are essential, what is the opportunity for vendors to differentiate?
IWB: Customer satisfaction, customer service, add-on services that they provide. Just because you have standards doesn’t mean your total offering needs to be standard. It’s perfectly fine that the basic offering is standard, and then if you want to lock people in, it should be because you are adding extra benefits that they wouldn’t be able to live without—value-added benefits. I think that will be the battle—the value-added features. But I honestly think the biggest value of all is how well you provide customer service. If you do a really good job, you’ll retain your customers. And if you do a poor job, I don’t think you will.
PwC: How far along are we in automating IT and data center operations? What challenges lie ahead?
EC: Automation can happen by different means. We can build hardware that is self-healing and autonomic. When that reaches a limit, we can build software systems that are so resilient that they can jump in when the hardware underneath fails. There are a lot of ways to go, but then consolidate it all for high availability. When you have 99.999 availability and you are the user, you don’t care how that is reached. Today, the best are at 99.999. We can go to 99.9999. When you measure it in minutes of unavailability per year for the whole system, that should be what ultimately matters.
IWB: I think it’s an Alice in Wonderland situation that we are running like crazy just to keep up, because in the meantime the workloads keep exploding. If you look at Smart Planet kinds of workloads, congestion management, infrastructure management, energy efficiency, and things like that, those are enormous workloads. To keep the system management needs from growing with the workloads requires massive innovation, so we are constantly running. And then new workloads will come up.
I think system management, autonomic management— I don’t see an end to them. Those are things that will constantly be innovated. We’ll keep making the hardware smarter, the software smarter, the management console smarter.
Also, the probability of security problems, hacking problems, reliability, self-configuring—there’s a whole set of things that we need to keep adding to make them easier to use, more reliable, and lower cost.
PwC: How will cloud computing change the mission and role of CIOs and their organizations?
IWB: I think cloud computing brings much more discipline to the CIO role and the CIO’s organization. The discussion would not be that everything becomes a cloud. The discussion would be about picking those services you offer that are growing the fastest that can benefit the most from being standardized and mass customized, and moving them to be more cloudlike. You can buy cloud systems and software, or you can have a company like IBM run it for you and have a plan to phase that over time. It’s possible that a lot of what they do will never move, either because they are more like back-end transaction services that are not cloudlike, or because they are legacy client-server things that are stabilized and the effort to migrate them isn’t worth it. So cloud computing imposes a discipline that’s very healthy.
To manage this discipline, you will need a very good CIO to decide what to do on-premises and off-premises, to coordinate the vendors, and to make sure the vendors are good. It’s like if somebody is manufacturing something for you and they use lead paint, it’s your problem. So you need somebody to make sure they don’t use lead paint and they do what they promised to do. And if you have IT suppliers, you need people to manage that very carefully and to interface between them and the rest of the business. So the IT function changes.
EC: When CIOs managed the centralized computing, distributed computing came in through the departments. Why? Because centralized computing was not responsive enough for everyone and departments shifted to a more flexible model. Cloud presents the same challenge and there is a possibility that departments will take matters into their own hands if flexibile, secure, and “sanctioned” options are not provided. CIOs can do two things. CIOs can ignore the growing interest in cloud computing, and that is not a good idea because the economics tend to be overwhelming and if you have overwhelming economics, usually sooner or later it’s going to happen. Or they can evolve into much more of an integrator of internal and external services, making sure that the integration happens in the right way. This is a very interesting and powerful role, but it needs to be actively managed.
Governance, in other words, is going to be key in this transition. If the governance is right, the benefits will be realized. If the governance is wrong, they will incur hidden costs of integration. Customers love the idea that they can swipe their credit card to provision some systems on Amazon.com, but they are telling us that the acquisition of these virtual machines needs to go through the right internal processes, to be auditable, and to be able to be monitored. If CIOs want the benefit, they need to integrate cloud activities with IT governance. Otherwise it will be a problem.
We see the biggest demand in helping customers adopt the model while still fulfilling enterprise demands. Individual developers can swipe their card and develop a little application for the iPhone, but if an IT project inside of a corporation deals with its IP [intellectual property], has a timeline, and has a group process, then you need a progress report, you need to have ITIL [IT Infrastructure Library]-certified processes, and you need to have a SAS 70–certified provider. Enterprise computing needs enterprise computing standards.
PwC: As your vision of cloud computing unfolds, how will your customers be different? How do businesses change as a result of cloud computing trends?
EC: They will look like services customers, so in many cases they will contract the outcome versus the how to do it. Second, instead of talking to me about lock-ins, they will talk to me about customer satisfaction and service level agreements. That is at least as big a transition for the company that delivers the service as it is for the customer. And we should not be surprised, because this is a maturing industry. In a maturing industry, the differentiation moves to the interface with the customer and the interaction is on results—not on how to do it.
IWB: I think businesses are spending way too much money and time on things that don’t bring them value, but there are things that they need to do, such as managing technology environments. As a result, they don’t have enough time to focus on the things that do bring them value. I think that’s why cloud computing will grow. Eventually everybody must move in this direction, because if more and more businesses are able to make the shift, then in principle they should do better than their competitors who are still wasting their time managing technologies. Eventually, everybody has to do it this way.