Remarks by Bob Muglia, President, Server and Tools Business
Bellevue, Washington
April 28, 2010
BOB MUGLIA: Good morning. It’s great to have a chance to talk to you today. I’ve often said hosting is a very important part of my business, because we’re really about building servers. And our mission is the servers and tools to power the world. And as we move towards the cloud, our partnership becomes even more important in the years ahead.
Now, as we look if you go back to early March, Steve Ballmer did a speech at the University of Washington where he talked about Microsoft’s commitment to the cloud. And he used this phrase, “we’re all in.” And he talked about five dimensions of the cloud. And the cloud has a very, very broad set of impact on the world as a whole, the way we all live our lives, the way we do business, and it certainly has a very significant impact on Microsoft’s business. I believe it has a significant impact on your business. And it certainly will evolve and change in, I think, a very positive way our relationship together in the years to come.
Steve talked very broadly. Today, what I want to do is focus in on the shifts that the cloud will bring to the cloud computing landscape, and just how that will change the way we work together to serve our mutual customers. I will tell you that when I look at the cloud, it’s one of those dramatic things. I’ve been at Microsoft 22 years, and I’ve seen some dramatic things over those numbers of years. But the cloud is right up there on all of the things that I’ve seen in terms of very large-magnitude shifts in the industry. Certainly the PC, the advent of the PC, was such a shift, the Internet was another shift in the mid-1990s, and I would put the cloud in the same category of impact in terms of the way all of our customers were building their business applications, and the way information is delivered to people, whether they’re within business or whether they’re consumers around the world. And I believe everyone in this room has a major impact in that. And we’ll talk a little bit about how Microsoft’s goal is to really partner with you to drive that forward.
Now, I say it’s a huge impact on the cloud, but it’s not just one of those exciting impacts, it’s also one that has a little bit of trepidation attached to it. Let me ask a question, is there anyone here who thinks of themselves as a climber? (Show of hands.) I see a couple of hands going up, not many. Not many people are climbers. Well, I wouldn’t say I’m a mountain climber, but I’m certainly a person who loves to hike, and occasionally when you do a hike you wind up in a situation where doing a climb, which is a bit more technical, becomes something that’s important.
You know you can reach the top. You know you can do it. You know, it’s something that’s exhilarating in the process, but sometimes when you get up there, and you’re sort of about halfway up, and you look down, and you go, holy cow, you realize that this is also a little bit scary. For me, at least, my palms get a little sweaty, not the best thing when you’re climbing a rock. But you’re in this situation where at one level there’s exhilaration and opportunity, and at the other level there’s a little bit of trepidation, and it’s a little frightening.
And that is exactly the way the cloud is. I wake up in the morning exhilarated by the potential of what this technology shift can do for our mutual customers. And the impact that we can all have on the world. And yet, it is a complete shift of the business model. It’s a shift in the way we work together. It’s a shift in the way Microsoft will have relationships with all of our customers. It will affect our business models across the board, not just in the context of how we work with our hosting partners. It’s broader than that. And so it’s a little scary at the same time.
But the thing that I want to emphasize is that this shift is happening. It’s an industrywide shift. It’s something that is happening all around us, and the technology has reached a point where it is going to have this transformative effect. And we all kind of face this opportunity to just sort of look where we are, and do a little naval gazing, or just really leap in with both feet and move forward on it.
I can tell you that my CEO is super clear. I totally agree with him. But we are all in on the cloud, and the recognition of the fact that this transformation is something that will be very positive for our customers. It will be very positive for our business. It will be very positive for our relationship with all of our partners, including our hoster partners, in the many years ahead. It is a big deal, and it is something that I think we will all profit with together.
Now, I do want to say a couple of things here. There are a number of challenges. Austen talked a little bit about how as technology shifts, businesses that have traditionally had strengths become less profitable over time. I think Web hosting is an example of that, very simple Web hosting, where we saw higher prices at the beginning, and we now see fairly low price offerings by a number of you, and we see Google out there at zero dollars.
So, we know that some things that at one point look really good over a period of time don’t look as good. And we have faced a similar set of things. When we build Windows NT in the 1990s, the kernel was a major competitive advantage. And our memory I mean, I used to sit up and talk about the memory management features of Windows NT in the 1990s. There is no value in the differentiation associated with that. Linux sort of wiped that away in the late ’90s and early 2000s, and there was a lot of prediction that Linux would wipe our Windows Server certainly by now.
If you go back to 2002 or so, the predictions were that Linux would be the dominant operating system in the world, but Windows today represents about 75 percent of all X86 servers that are shipped on the planet, and over 60 percent of the revenue that’s earned by the entire ecosystem. So, how does that happen? Well, the reason we’ve been able to profit and grow against a free competitor is because we recognized that there was customer value that could be delivered through differentiation that was not being supplied there. And some of that is delivered by Microsoft, a lot of it is delivered through our partners.
We know our partners really make on the order of $7 to $8 for every dollar that Microsoft makes. And as we sort of look forward to the future I expect that to be true. And this change that the cloud is driving, where some things that used to hold value will not be the place where we profit, that’s true for us, it may be true for you, we just have to face that. And we have to drive forward with the recognition that the cloud will bring with it a whole new set of opportunities that we can all capitalize on.
And I want to say one last thing before I exit this slide talking about the cloud is that, none of this is to say that non-cloud traditional datacenters will go away. This speech is all about the cloud, so I’ll put all the rest of my focus on this, but I want to say that we sell a lot of Windows Servers and SQL Servers into the market, and customers run them in traditional datacenters. They’re non-virtualized. Virtualized servers are only about 20 percent of all servers shipped right now. And we expect that there will continue to be large numbers of traditional servers sold.
So, the old business model, the previous business models that we’ve all profited on don’t necessarily go away. Let’s face it, there’s still a lot of money that IBM and the mainframe ecosystem still makes on mainframes, 30 years or so, at least 20 years after its supposed demise. That market still exists. And so there’s still a significant opportunity in what we’ve all become accustomed to. But, we’ve kind of come to the conclusion that we can’t operate that way.
We could cling to what we have right now and say, hey, we got this cash cow, let’s just milk it, and we’ll do a little milking. But, mostly what we’re going to focus on is our energy. Most of my R&D is shifting, or has shifted, to thinking about what it means to do the cloud. So, let’s talk a little bit about what our approach is. In a lot of senses our approach with the cloud is exactly the same as Microsoft’s approach has been for as many years as you can remember. Microsoft is it’s an old dog in some senses.
It’s well known. We’ve always had a consistent playbook and our playbook for the cloud will be consistent. It will remain, understand our customers really well, build a highly differentiated product, price it very cost effectively, sell it in volume, and work with partners. That’s page one in the Microsoft playbook. I’ve given that speech a lot of times in my career. I’m saying it now that that’s exactly the same thing. I learned that, by the way, back in the 1980s, and that’s what we were doing then, that’s what we did in the 1990s, what we did last decade, and as we move towards the cloud we’ll continue to do that.
And so our focus will be on how we build the software that can power the overall ecosystem towards this next generation, that remains our number one goal. If I look and say, we have 75 percent of all X86 servers, hey, I want 75 or 80 percent, my objectives are to grow my share as we move into the cloud world, and there will still be heck of a lot of servers that are sold in this new world. And I continue to believe that the server market will continue to grow. But, a lot more of it will grow in the form of cloud-based services that, frankly, will be provided by the partner community, our hoster community, and that’s delivered there.
So, how we work and deliver our technology is very critical. Our playbook won’t change as we go into the cloud. The specifics of the value we deliver, the specifics of the value you deliver, some of that will change, some of that will change going forward, but we’ll continue to provide this highly differentiated, very horizontal platform that our customers will see differentiated advantage associated with using, and that you can leverage and use in your business to provide differentiation to your customers, and really provide you with the opportunity to grow your overall business.
Now, one thing I will tell you about the cloud, and I’ll talk a little more about this when I get into the details of what the cloud really means. One thing I will tell you is that, I’ve been working with providing software to hosters for well over ten years, and we have not done as a company as good a job of providing you software that was designed to run in your environment, and thus you’ve had to do a lot of make work to adapt it to work.
Let’s go back. Think about a few years ago before PowerShell existed, and the lack of automation capabilities that were in Windows Server, and how Linux was actually quite a bit better than we were. You know, we recognized this was a deficiency, that’s why we did PowerShell, which I think is quite differentiated today, and it provides a set of underlying automation services to make it easy to run datacenters at scale. Well, part of the reason for some of these deficiencies has been that our focus has always been towards enterprise customers, and small business, not so much delivering things as a service, and particularly not so much towards the hosting community.
And one of the things that I think will help us all in terms of where we work together is it turns out that as we all move to delivering services to our customers today we build software that we sell, tomorrow we’ll build software that is delivered as a service that our customers will acquire and use on an ongoing basis. That’s one of the attributes that comes with the cloud. But as we move to that, our focus will shift to do things that are much more compatible with what your business needs are, and allow you to focus more on differentiated value that you can drive rather than fixing the deficiencies that exist within our software.
So, our approach stays the same as we move towards the cloud. The opportunities are different, and the specifics of how we are able to provide value to you will change, and I think it will change in a way that you will find to be very, very productive.
Now, one of the things I want to do is say that we have done a lot of work to date. There’s much ahead of us in the cloud, but we’ve done a lot of work to date with our existing products, Windows Server, System Center, SQL Server, as well as adding value in terms of the Dynamic Data Center Toolkit to enable our partners to be able to build clouds today with our existing products. So the reality of moving into this direction is very real today, and there are well over 40 hosters around the world that are actually doing that today on top of our existing products.
What I want to do right now is invite Hyder Ali up to show you an example of how one of our hosting partners is providing very differentiated value on top of today’s existing Windows Server, and System Center platforms.
Hyder.
HYDER ALI: Hey, Bob. Good morning.
Today, I’m going to demonstrate a disaster recovery solution that we have built on top of System Center Virtual Machine Manager using the Dynamic Data Center Toolkit. What you see here on the screen is the customer view of a control panel from a hosting provider. You see a couple of VMs. If I hover over this VM, you will see that this particular VM is disaster recovery-enabled. If I were to connect to this VM, you will see the name of the server in which the VM is hosted on. You see the name, you’ll see that it is HV-Host02.
Now, I have a colleague of mine, Mark, who is standing backstage, and he’s going to simulate a disaster. And he’s going to do so by creating a tornado in the Midwest, and he’s going to do that by simply bringing the hard disk on this server offline, which is going to cause the VM, the application to come down, and we’ll seamlessly move the application and the VM to another datacenter.
This is happening live from the datacenters of Engine X. Engine X is a managed hosting provider in the United States. And let me dive behind the scenes to show you what is happening. To my left, and to your right is the primary datacenter that is located at Evansville, Indiana, and our secondary datacenter is located at Milwaukee, Wisconsin. And these two datacenter are about 500 miles apart, or 800 kilometers. What we have running on these two datacenters are two clusters of Hyper-V that we built using Dynamic Data Center. We also have a customer application running on a virtual machine in the primary datacenter. What is also unique about this setup is that there are two NetApp storage servers at these two datacenters that are actually doing synchronous replications.
BOB MUGLIA: So, in this case, what’s happened is, Engine X is built on top of taking the technology that we have today with Windows Server, and Windows Server Clustering, as well as System Center, and then building on top of a solution that we have available Dynamic Data Center Toolkit, and provided a custom user interface for their customers to make available virtual machines that have a set of differentiated capabilities particularly here is high availability. In the case of a datacenter failure, or a hardware failure within a single datacenter, or even a communications failure to a datacenter, the system is designed to be able to replicate over, and continue operation, providing business continuity out of a secondary datacenter. This is an example of something that’s extremely expensive for customers to do today, and a great opportunity for the hosting community to provide differentiated value on top of our existing platform.
Now, I say that our focus has always been to work with the ecosystem to provide this full set of solutions. I mean, in this case, we’re obviously working with Engine X that’s providing the solution to their end customers. We’re also working with other partners. Engine X has chosen to use NetApp, which has this high-availability storage solution, and they’re providing the underlying storage replication between these two datacenters, which are pretty far away, almost 500 miles away. And, they’re providing a core set of the technology that Engine X has built.
Now, we’ve worked with NetApp, work with other storage partners as well, but NetApp does a great job. We’ve worked with them to provide to enable these integrated solutions to be delivered by our partners, our hosting partners, as Engine X has done. And there’s a lot of value in this. This is something that customers are willing to pay for, and really and it’s thought of as a fairly differentiated thing.
HYDER ALI: So, what you see here, if we were to switch back to the customer portal, what you see here is a portal. In this particular case, the portal was built by a control panel partner of ours called EMS-Cortex. Engine X worked with EMS-Cortex to customize a portal for their customer needs. And what you see here is that you see an alert that showed up here, and if I were to click on this alert, you would see an error message would show up that points to the fact that a disaster occurred in the primary datacenter, and we seamlessly migrated the applications to the secondary datacenter.
What’s happening is that there is an operations manager that is monitoring both datacenters, and when it detected a failure in the primary datacenter, it created an alert, and it ran a script to initiate the startup of a VM in the other datacenter.
BOB MUGLIA: Now, what Hyder just said is important, because one of the attributes that we’re providing broadly to our enterprise customers as well, of course, to our hosting partners, is a very broad platform. It certainly consists of the underlying virtualization technology in the operating system, but it also consists of a very complete management solution in the form of our Virtual Machine Manager technology, our product there, as well as Operations Manager, which can manage the status of applications as well as the underlying hardware environment. And so, really, that complete end-to-end solution, Microsoft is very unique in terms of what we can deliver, and it’s achieving a very high adoption rate in customers.
To give you an idea, in the enterprise, roughly 50 percent of all servers that are shipped into the enterprise are running System Center on it today. So, we have incredibly high adoption, and that’s across virtual as well as physical machines. We have a very high adoption rate, and a very high degree of customer understanding and acceptance of this environment. Again, that provides more opportunity for you as you think of providing differentiation.
HYDER ALI: What you see, the server on which this VM is hosted, you will see that it is a different server that it is based out of. You see it’s BRK-Hyper-V. So, what is interesting about this scenario is that the customer of you, of the hoster, as well as the end user are transparent to this disaster that happened. And in this particular case, Engine X offered the service as a service to their customers in the same way that they would offer other services to, like backup and restore.
So, this is a solution that we’ve built, we’ve demonstrated today through partnership with EMS-Cortex and NetApp, and I hope that you’ll be able to leverage this to your own customers based on our technology.
BOB MUGLIA: Thanks a lot. (Applause.)
The key point I want to say is that I’ll spend the rest of the talk talking about sort of where we’re going in the future with the cloud and what we’re delivering overall and the way we view the cloud. But, there’s an enormous amount of technology that can be applied towards building clouds, clouds for your own environment today to deliver to your customers, be they private clouds, or public clouds. You have the ability to do this today with the existing technology, and we applied a number of partners that are actively doing it, and quite successfully to boot.
So, I want to talk about the cloud now and really talk about it’s cloud computing, it’s the aspect of the evolution of the datacenter towards a compute an overall shared compute environment. And I thought it’s useful to define what a cloud is, and talk about the implications that has on all of us, and our customers. And I tried to take the simplest definition of the cloud that one could imagine, of cloud computing that one could imagine. And there’s a lot of attributes associated with cloud computing that are also value add on top of it. But, the underlying essence of cloud computing is really doing just-in-time provisioning and scaling of services, software and solutions delivered as a service on a pool of hardware, a set of hardware that all works together in a consistent way.
That’s really the essence. There are many attributes that the cloud can bring to applications, always available in business continuity, like Hyder just showed. That’s a great attribute. Elastic scaling of an application, that’s a great attribute. Federation of identity, another great attribute. Many, many things that are important to deliver as a part of a cloud platform, but the essence is really this idea of just-in-time provisioning and scaling of service on a set of pulled hardware.
Now, the implication of this is that we see this as an opportunity to dramatically accelerate the speed of building solutions and delivering solutions into the marketplace. I think that’s true for your own environment and as you think of the solutions and offerings you want to have for your customers, I also think that’s true for our end customers that are really looking for a differentiation that they can get in the market by building applications faster and at less cost.
And the costs will be driven down associated with that, and that provides some interesting challenges, but it also provides some great opportunities, because the cost of building out very, very powerful solutions will be dramatically less than what is traditionally available in today’s world.
Now, when thinking about this, there’s often a conversation, there’s a lot of conversation out there around public and private clouds, and our customers have interest in thinking about both. We see services like Amazon Web Services or our Windows Azure, which are public cloud environments by almost anybody’s definition. And when I talk to CIOs, I had a number of CIOs in last week into Redmond, and these were large U.S. customers. And almost every one of them had an interest in building what they thought of as the private cloud. Now, I do think that these definitions lack the clarity of what customers would want to do, because there are more attributes that are interesting to understand.
So, in honoring and recognizing that the industry is using the term public and private, we’ll continue to talk about it in that form. But, I want to provide you a click stop down on the way we think about what sits behind building these so-called public and private clouds. And when we consider this, we sort of look at two dimensions. What I’ll call shared and dedicated clouds. A shared cloud is a cloud that supports multiple tenants. So, on this pool of hardware there are multiple organizations that are acquiring and using services. Then there’s the dedicated cloud, where you have set of hardware that is providing a set of services for a single organization, a single enterprise customer, for example.
Both of them are interesting. Both of them turn out to be very interesting. There are many cases where customers want what they would think of as a dedicated cloud, where they are very clear what the hardware is, and they are very clear that they have control over what’s running in that environment. And there are also many cases where the economics attached to a shared set of underlying hardware that’s provided across multiple organizations is very interesting.
Now, when we think about this, we think about how these clouds, shared and dedicated, will run in customer datacenters, in our partner, hoster datacenters, and in Microsoft datacenters. And when we think about this and we see those six cells in that matrix, all of them turn out to be interesting. They’re all interesting cells to think about in terms of providing services to our customers.
Some are, I think, fairly obvious. I am very confident that many of you, or maybe over time most of you, will build up a set of cloud services that you will offer to your customers. And many of you will provide those, both in the form of what I will call a shared cloud, where multiple organizations are sharing a set of underlying resources you’re providing, and you will provide dedicated services where customers will really know the hardware that they’re providing. So, in the partner space the fact that there’s two checkboxes there is hardly surprising, and again, we want to provide you with the underlying software capabilities to enable you to most effectively deliver a differentiated solution.
On the customer side people go, well, wait a minute, aren’t all customers doing dedicated clouds? And it’s probably true that most customers, when they think about a private cloud, they are thinking about a dedicated cloud that runs in their datacenter. But, there are cases where I talked to customers that are interested in providing what I would call a shared cloud that runs in their datacenter.
You know, an example of this are financial institutions that provide services to other financial institutions, where they’re building a piece of software that they’re essentially it’s a product of theirs, this financial software is a product that they are supplying to other financial services. And they want to run that inside their datacenter. They are usually pretty clear about that. But, they want to allow that software to run on behalf of another organization, another bank, so to speak. And I see it in a few places. You can think about it in dealers, in auto dealers, as an example, where customers would want to run the datacenter environment themselves, and have control over the underlying hardware, but they would do it for multiple organizations.
Now, in the case of Microsoft we will mostly when we think of Windows Azure, it is a public cloud environment, and the vast majority of energy will go into thinking about it as a highly scaled public cloud, and shared cloud environment. But, it’s not the only case. There are at least some cases where that where we need to think about providing a dedicated environment. Perhaps the best example of this is government organizations, maybe not the only, but the best example is government organizations, where almost every government wants to participate in some form of having what I would think of as a dedicated cloud. So, we’re having conversations in that regard. But, most of our focus will be in the shared environment.
That’s sort of a quick stop down and a way of thinking about it, and our strategy very simply is to provide a consistent platform, one platform, one application model, one management solution that spans all three of these.
And so the implication of this is that the learning and the technology that we’re getting from Windows Azure, and as we build out the Windows Azure environment, we’re bringing back into the products that we’ll sell to enable our customers, whether they are enterprise customers or yourselves, to build clouds in your own environment, and we’re also talking the learning that we have in Windows Server and System Center, and SQL Server, to bring all of that forward into Windows Azure. It’s a very bi-directional set of technology transfer across those things.
Over time we will converge those offerings into something that’s very consistent, and is available to our end customers, our hosting partners, and we’ll run ourselves. And I’ll talk a little more about how we think about doing that. So, this vision, simply put, is a consistent environment across all three of those dimensions, which enables dedicated and shared clouds to be built with common identity environment underneath, common management tools, and common dev tools.
Let me tell you, when we think about building a cloud, there’s a lot of changes associated with what that means. The software that we’ve been building for many years is really not cloud-ready. And we have been investing in making those changes. Let me give you a couple of really key examples. Federation and identity federation, we will live in a world where identities are authenticated by authenticating services that exist in different spaces, in many cases it will be the active directory that is running on our customer premises.
Sometimes it will be authenticated within your datacenter, sometimes it will be authenticated in our datacenter. We need to enable a federation so that those identities that our customers wish to have can be used across services, regardless of where they’re supplied. A great example of a software capability that Microsoft is committed to delivering to our enterprise customers, as well as to our hosting partners, and we’ll run it ourselves.
Another really good example of how the technology changes is in multi-tenancy. Our software historically has not been built to be multi-tenant. It’s not been designed to run in that shared environment. We’re going to change that. I mean, everything has got to change. Everything has to be designed to be multi-tenant. Sure, there will be dedicated cases where this isn’t important, but the software has got to be designed to be multi-tenant.
And you guys a lot of people have run Exchange and a lot of hosting partners have run Exchange, and of course we know that historically Exchange was not really designed to work in that environment, so there was a lot of work that you needed to do to essentially spackle over that. And we’re investing in our core products so that those capabilities are fundamental within them, so you can put your energy on providing real, differentiated value to customers that are unique, and that only you can really provide.
And the key to this is that we’re doing this in the context of all of the solutions, all of the products that you’re familiar with and running today. So, the investments that you’re making today in hosting Windows Server, using Hyper-V, working with the Dynamic Data Center Toolkit to build cloud platforms in your own environment, those investments will be carried forward. The PowerShell scripts you’re writing. All of the sets of things that you’re doing to take these widely used products, and bring them into your environment, and offer services to customers, those investments will carry forward.
Now, we have a lot of work to do in terms of bringing the technology back to you. And I’ll talk a little bit more about how we think about that, and how we plan to do that, because the cloud, in addition to providing changes like federation identity, and multi-tenancy, there’s a whole different way that we think about running this pooled set of resources. And we’ve learned a lot by running this ourselves. We’ve learned a ton by running this ourselves, and we’ll be delivering that to you.
So, in talking about that, I mentioned earlier that in our enterprise products, we learned by running these things ourselves, and we’ve never have the equivalent in the cloud environment. You know, MSIT is our IT organization. And we have over 200,000 desktop users. There’s not that many people that work at Microsoft, but with contractors, and all of our extra desktops, and stuff, there’s over 200,000 desktops that run in Microsoft in tens of thousands of servers that our IT department runs. We are a superb example of a large, geographically distributed enterprise customer, and we learn a ton by running the datacenters ourselves, and running the environment ourselves. We simulate that. We dog food it. We call it dog fooding. We dog food it, and we fix a lot of problems. We learn a lot. We get a lot of requirements in before our customers ever see the product. We’ve never had that same environment for our hosting partners. And, as I said, there are some deficiencies attached to what we’ve delivered to date.
Windows Azure provides us with an environment where we do that learning, and we learn a lot by running Windows Azure at scale. With Windows Azure what we will do is provide a very consistent platform environment that has a well-defined set of software capabilities run at scale that is, in fact, geographically distributed. And our energies there will be focused in on providing software platform value that our ISVs can take advantage of, and our end customers, our enterprise customers, and small-business customers, everybody, all of our customers building apps can take advantage of.
And also understanding how to operate and run this at scale, we’ve got to do it ourselves. We just learned this. It’s one thing for us to build something and throw it over the wall, and you guys figure out how to make it work. It’s another thing to live it every single day, and understand the pain that it takes to run and operate these things for end customers. And that learning will translate into what we can deliver to you in the software and services we deliver that you can run inside your environment.
Now, what we’ve done with Windows Azure is, we’re building a platform as a service. We think of it as a broad platform that people can build applications on, and it has these cloud attributes. It is a complete platform in the context that it provides the underlying operating system environment, which includes the Web tier attached to it. It includes the database. And it includes the middle tier that provides the middle-tier services.
And so we think about how, as we move towards the cloud, we can take all of the learning we’ve had for 20 years, and build in what we think of today as Windows Server 2008. We can take that and bring it forward into the cloud environment, and there are major differences. There are differences in the way we think of, for example, running SQL capabilities. Our SQL Azure capabilities today that we’re delivering is not a single SQL Server that we run. It is actually a fabric-based environment of well over 1,000 database servers spread across six datacenters globally.
And we’ve focused on how we can make that self-managing across those things, so that the creation of a database inside that environment is an automated step that is done, that can be provisioned by a customer. The idea of how those databases are maintained and replicated is all driven automatically.
And one of the words you will hear more of in the coming months and years is “fabric-based.” And the reason we called our middle-tier services AppFabric is because we’re building out that middle tier on top of an underlying fabric system that’s designed to take the resources of multiple virtual machines, five, ten, hundreds, thousands of virtual machines, maybe even tens of thousands. And taking those resources and treating them as a consistent environment that can provide a set of services for applications.
Windows Azure is built on a fabric, it has a fabric controller built into it. Our middle-tier services that we’re providing in this environment with AppFabric are built in this environment, and SQL Azure is also built on a fabric environment. And those concepts, all of those concepts, we are looking at how we take and make those available and bring them into our core products that you can acquire, and that our customers can acquire as you think about implementing cloud.
Now, there’s an interesting process here, and this is a major sort of shift in the way we think about this, because we’re learning in the cloud. We have a piece of technology today that we’re thinking about like some of these underlying fabric services. We design it. We develop it. We test it. We deploy it into the cloud in a beta environment, we’re testing it now. We’re dog fooding it ourselves in our own environment. When it stabilizes, we can release it today in the Windows Azure environment in production. Over time, we want to be able to continue that staging, so we can provide those same set of services on a much more consistent delivery mechanism to our enterprise customers, and to our hosting partners.
I say this is a big change. It’s a big change. It’s a big change in the way we build software. It’s a big change in the way we deliver software as a service to our customers. And I do find it both exciting as heck, but also kind of scary at the same time. So, Windows Azure is an environment for us to learn, so that we can deliver to the breadth of all of the people that will consume this software as a service.
Let me talk a little bit about the implications of the cloud, because I think it’s incredibly important to understand the impact the cloud will have. And when I talk about the cloud, I talk about it as having three this is cloud computing, excuse me. When I talk about cloud computing, I talk about three major model shifts that occur. And those model shifts are the hardware model, the application model, and the operations model. And each of them is incredibly dramatic in terms of the impact.
Let’s talk about the hardware model for just a second. We’re looking at how as we acquire servers, how we work with our OEM partners to change the kinds of products they’re building to deliver products that provide absolutely the highest performance at the lowest cost with the lowest power usage. Those are the three attributes. We say to all of our hardware partners, you want to bid on Microsoft datacenters, and by the way we’re at this point, Google builds more servers than we do, but they don’t buy from our OEM partners, the same folks you buy from. We do. And we’re now the largest purchaser of hardware on the planet from OEM partners. And our goal is to help those partners understand how to deliver the most effective hardware at the lowest price, and lowest power usage. And those three attributes together are incredibly important.
We, a long time ago, 15 year ago, when we started doing MSN, we used to buy servers in boxes, and we’d take them out of the box and put them in a rack, and wire them up, and test them and run them. About five-six years ago, we stopped doing that, and we bought racks that were all preconfigured from our OEM partners, and the racks would come in. In the last two years or so, it’s really been growing very rapidly in the last 12 months, we stopped doing this, and we’re now buying containers.
And that’s a picture there of our Chicago Datacenter, and some containers that are running. I think those containers are running Bing and/or Windows Azure. And we’re now buying 2,000 servers at a shot in containers, and we’re working with our hardware partners to say, well, how can we ensure that we get the most computing power at the lowest cost, and the lowest power consumption? There’s a lot of innovation that we can help to drive into our hardware partners.
We see this as having a potential impact in terms of reducing the cost of hardware per CPU, per performance, and per gigawatt used by a factor of 10. So we think there’s a 10x improvement in the cost and making these things much more efficient. And we are working with the HPs, and the Dells, and the SGIs, and others in the world that supply this technology to ensure that this is available to you, and whether that’s a whole container which has about 2,000 servers in it, by the way, which for some of you it’s very reasonable to think about purchasing at that level, or perhaps a fraction of a container.
Certainly, when I talk to enterprise customers they’re more interested in buying clouds that have a couple of hundred servers at a time. But, providing that same level of cost reduction to everyone, to the entire ecosystem, and Microsoft is very much helping to drive this across the industry. So, we think there’s a factor of 10 there.
In thinking of an application model, the key is how do we help our customers and ourselves, frankly, build solutions that differentiate our business. Faster, it’s all about how fast can you build an application and bring it to market. And when we look at some of the attributes that are coming in cloud computing, the impact that modeling will have on the way applications are designed and written, and modeling will provide an understanding of the characteristics of an application, and move from procedural code to a declarative language, a significant part of the application logic.
That, combined with this underlying fabric system that will provide a basis to do a scale-out implementation that spans across a large number of computers, and provides characteristics like stage production to enable always-available applications that scale elastically, based on demand, those characteristics are inherent in new applications that are built for the cloud. And we see a tremendous opportunity to speed the time to market, getting applications to market a lot faster by taking the technology that we understand how to use in building applications in this cloud environment, and applying it very, very broadly.
Now, one of the attributes that all of us face, certainly our own code, we face this a lot, is our existing applications were not written for the cloud. And one of the things that we’re in conversations with all of our customers with is how do you take an application and move it into the cloud. And this is a complicated conversation, because there are some applications that are so highly intertwined in a customer’s environment that it’s tough to move into the cloud. I don’t know how to move one of those mainframe applications to the cloud. I really don’t, except rewriting it.
But, there are a number of applications that will benefit from the cloud, and certainly at an infrastructure level, if you move an application as it exists, package it in a virtual machine, and move it into a cloud environment, you get the benefit of the reduced hardware costs, and then the operations cost that I’ll talk about in a second. But, that’s not really getting the full benefit of the time to market. And existing applications can be evolved to take advantage of some of these new characteristics, based on the business need of that application. If it’s a legacy application, and it can be moved at the infrastructure level, at the virtual machine level, great. Take advantage of the reduced hardware cost. It will run nice and happy up there. It won’t be elastic. It won’t be always available, but it will have a set of characteristics that it’s always had and it’s cheaper to run. It’s fantastic.
Other applications are evolving and are important to the business, and it makes sense to take advantage of some of the characteristics of the cloud, such as enabling business continuity, always available, scale-out elasticity, those are attributes that can or federated identity, those are attributes that can be incrementally added to existing applications where it’s worthwhile doing it.
But, not all apps meet all characteristics. Not all apps need to be multi-tenant. Many of our customers don’t care about that. So, there is a number of cloud application attributes that depend on the application. Our goal is very simple here, to do what we’ve always done, which is to take the knowledge associated with building this next-generation set of business applications, and bottle it, and enable it, and make it available to all of our customers, and to all of you.
There are a number of companies that can build clouds today. I mean, if you look, Microsoft has a cloud, a big scale-out cloud. Amazon has a cloud. Google certainly has a cloud. A number of you are working with us to build clouds, and we want to continue to do that. But, our goal, simply put, is to take the knowledge that we have as we build all these clouds and transfer it to our entire customer based and our partner base, to enable all of you to profit from it. It’s what we do. It’s what we’ve done for as long as we’ve been in business. It used to be hard to build Windows apps — Visual Basic made it easy. It’s hard to build cloud solutions today. Microsoft products, Windows Server, our products as they evolve will make it easy for you, as well as for your end customers.
The final piece of the operations model, the final piece of the cloud shift is the operations model. And here is a fundamental shift in the way you think about running servers at scale, and we know that taking traditional applications and running traditional applications in a traditional way has issues and costs associated with it when you try and scale it up. Our customers and I think many of you, and certainly ourselves spend a very, very high percentage of your overall dollars in running an operating system, and the people cost associated with running an operating system.
Another story, when we built our MSN 15 years ago, we built it traditionally, just good old traditional applications, using traditional IT technology. So, taking Web servers and building a little middle tier, originally it was in C++, now it’s in .NET, hosting SQL Servers, et cetera. And the cost associated with scaling the operations was fully a function of the number of servers that we had. With our customers it’s not unusual for our customers to have an operator assigned to every 30 servers, and world-class enterprises run at maybe 10 times that.
When we run Bing, Bing has a couple of hundred thousand servers in it. You can’t have an operator for every 30 servers, or even every several hundred servers. It just doesn’t scale. We spend all of our money on the operations side. But, with Bing what we learned as we built it out, we took and we said, OK, there is a different way to operate where there is an assumption that servers will fail, but the system needs to be always available and resilient to those failures regardless of what happens.
So, in the Bing environment, several hundred thousand servers across multiple datacenters, failures are things that just get logged as trouble tickets that need to be solved. A failure of a piece of a server is something that needs to be fixed the next time a tech is in the datacenter and replacing hardware. It’s not something that affects the operation. The server is simply taken out of the overall fabric environment and then when the hardware is replaced it rejoins the environment.
That knowledge is incredibly important, and that’s what we’re learning as we build out our cloud services, and that’s exactly what we want to deliver to you, so you can have a much, much reduced cost of running these systems at very large scale. And just like the other things, we think there’s a factor of 10. We think there’s a factor of 10 in terms of reducing the costs to operate these systems. And this is learning that we’ve gotten. We had that learning in Bing, but Bing was not written in a generalized way that could be leveraged by generic applications. It was all done in a specialized way.
Windows Azure has been all about taking all of that learning and putting in a form that we can use more generally as a platform, and then as we move forward we’re taking this technology and we’ll be delivering it to our enterprise customers, as well as to our hosting partners.
And one of the things about this is that when you’re running thousands of servers, thousands, and thousands, and thousands of virtual machines, you sure as heck better not have thousands and thousands of images that you need to manage. That’s just one of those things that kills you on a scale perspective. You need to think about how the images that exist underneath are fully separated from the applications, and those things are all composited, or brought together as a part of a fabric-based runtime.
And with that what I’d like to do is bring up Edwin Yuen who will show you some of the things that we actually demoed this last week at the management summit that we did, the management conference we did in Las Vegas, as a part of where we’re going to with System Center, and some of the new features that are being built right now as a part of System Center.
So, Edwin.
EDWIN YUEN: Good morning, Bob.
So, today we’re going to give a sneak preview of some of the new capabilities in System Center datacenter. Now, abstraction is a key element of building cloud environments, because it separates the infrastructure from its capabilities. But, the real business value is being able to deploy those capabilities across different infrastructures.
So, what we see here is actually the next version of Virtual Machine Manager. Now, in existing versions of Virtual Machine Manager we’re managing virtual machines, and we’ve also had the ability to manage libraries of objects. What we’ve done here in the next version is manage server application virtualization packages, SQL DAC packages, and MS Deploy packages. Now, the SQL DAC package has the capability to configure a SQL database completely. And the MS Deploy packages configure the IIS systems. And with server application virtualization we’re able to abstract the application from the operating system and dynamically deploy that application right into a virtual machine.
BOB MUGLIA: So, I mean, literally what we do is, instead of thinking about the one virtual image as being the way you have to deploy an application, we have a smaller number of operating system images, and then those operating system images, when they’re deployed into production, the application is composited, or added to that, and then the whole thing is initialized.
EDWIN YUEN: Well, in the existing version of the MM, we’re using templates to deploy virtual machines for applications. But most applications are really defined at a service level, and span multiple virtual machines. Now, there’s been a lot of discussion about managing applications at the service level. Let’s show you how we do it in System Center.
What I’m going to use now is what is called the Service Designer Feature in Virtual Machine Manager. And it allows us to model, design and deploy applications at a service level. And in this case, I’ll do a scale out, three-tier application, which I call Contoso Three-Tier Template. And when I go ahead and enable that, you can see right away that the skeleton for the entire system, including the Web tier, the app tier, the database tier, the networks, the storage, everything is already laid out for me.
In the bottom left-hand corner, I have templates. Now, templates in this case are not only virtual machine configuration and an operating system image, but also an application configuration that’s associated with this. So, what I’m going to do is, I’m going to look at my Web tier, and if we look at my Web tier, we not only have the hardware configuration and the operating system configuration, we have that MS Deploy package, which configures the entire IIS Web site, including the site name and the physical path.
And to deploy that, I simply drag and drop it. And once I’ve deployed it, I want to make it a scale-out application, so I can say scale out. I want to make sure that I have a maximum of ten servers, and a minimum of three servers. We’ll go ahead and deploy the application tier, and with this application template, we’re using that Server App-V, or Server Application Virtualization image. And instead of actually going ahead and deploying the application after the creation of the virtual machine, or even running scripts to configure it, we can dynamically deploy that application in real time right to that virtual machine. And since the application and the OS are abstracted from each other, you’re going to be able to compose multiple, if not hundreds of, different applications using one OS image.
BOB MUGLIA: Now, this idea of image composition is something we are doing today with Windows Azure. And then Edwin showed a second ago how you can scale out from, say, three servers to up to ten servers, and the Web tier is designed to elastically scale. That, again, is ideas and technology that we first are pioneering and learning about inside Windows Azure. But here, with our products that exist today, and in the future here with Systems Center, we’re making these same things available very, very broadly.
EDWIN YUEN: And then we’ll continue with the composable images by looking at the database tier. And we have better or more powerful hardware. But, in this case, we have that DAC package that has the instance name, the database configuration, everything we need to deploy the database with.
BOB MUGLIA: So, OK, let’s take a second here. The SQL DAC is a set of configuration data essentially defines a model for a SQL Server database, which includes the configuration of the tables, and the views, the storage procedures, even the data associated, the underlying initialized data.
Now, in the past, what’s been required is to use SQL scripts to do this. Here it is a set of declarative configuration data that is all done inside the SQL DAC. Now, this is a case where the pioneering work is being done in SQL Server and will move to Windows Azure. So, we just shipped SQL Server 2008 R2. It has this DAC capability. It’s something you can use today. This feature is not in Windows Azure yet. It will be.
The idea that all of these things are crossing is where we’re heading. And this is a case where a very important capability that lowers the cost associated with deploying and maintaining SQL Server databases is available first to our enterprise and our hoster partners.
EDWIN YUEN: So, we’ve gone ahead and modeled the Web tier, the app tier, and the data tier. And if I were to press the deploy button, we’d actually go ahead and take that model and actually deploy the entire service, the virtual machines, and all the connections. But to save some time in the demo, we’ll go to an existing one that I already have, and we’ll switch over to the services view. And if we take a look at the view, we have this Web store service, which is actually what we deployed. And as I open that up, you can see it has the five virtual machines, the three from the Web tier, the app tier and the database tier.
Now, making it easier to deploy applications is important, but in the life cycle of an app, you really only deploy once or twice, but you’re constantly maintaining and patching systems. And we’re going to make that easier also.
So, if we go back into the library, we saw we had multiple services, multiple virtual machines, all of those systems are built using only three operating system images, a Web image, a SQL image and an application image. And in VMM now, we can actually do compliance scanning, and we can build a baseline that either the hosting partner or the customer can request. And when we go ahead and run the scan, it’s going to work with a W-sub server, look at that OS image, and then check for validation and any remediation that’s required. It’s doing all of this offline without affecting the running systems.
And then, as soon as it’s done, you’ll see that two of my images are fine, but my app tier requires remediation. Normally, I would go ahead and validate this, and test it out. But for the demo, we’ll go ahead and just remediate it. And now it’s applying the updates, again, offline to the image that we have stored in the library. But since Virtual Machine Manager is managing all the services, all the virtual machines that might be associated with this image, we’ll know instantly what other systems need to be updated, and we can go ahead and update those systems with a click of a button.
BOB MUGLIA: So, what we’re doing is, we’re taking a situation where a system needs to be patched, we’re doing the analysis to understand which images require the patch, and then we’re able to apply the patches all offline, not on the production systems.
EDWIN YUEN: Exactly. And once we’ve patched that offline system, we can see all these systems need updates, and we can see that this app VM requires the update itself. And all we have to do is go to the bottom here, and I press update service. And when I update the service, what VMM is going to do right now is, it actually takes that server App-V image, it extracts the application including the current application state and configuration off the operating system image, we remove the old operating system image, deploy the newly patched operating system image, repersonalize it for the app, and then drop the application right back on top, and it allows us to go ahead and actually update multiple hundreds, thousands of services, and applications simply with a simple button click by updating that one image.
BOB MUGLIA: I mean, this is something we learned in Bing. When you’re running hundreds of thousands of servers, you do not apply patches to all of those servers, because you have a number of them that have problems associated with the configuration. So, what we literally do is we redefine the images. In the case of Bing it’s relatively manual, we redefine the images. And then we redeploy groups of servers with the new sets of images.
Again, we have other groups running the old images in a mixed environment for a period of time, as the deployment is staged across the different clusters of servers. The application always stays available. Bing was designed to be able to do that. But, the images get reimaged on the the new images are put on each of the different servers that are associated with that, and brought back into the active cluster.
That’s exactly the technology, the learning that we’ve done and exactly the technology that we’re bringing into Windows Server and System Center.
EDWIN YUEN: Great, thank you, Bob. (Applause.)
BOB MUGLIA: So, the underlying point that I want to sort of leave you with is that the cloud is a very major shift. It is as big a shift as I’ve ever seen and it brings with it incredible opportunity, that is unbelievably exciting on the one hand, and presents a few challenges on the other hand. It challenges what we do today. It challenges our business model in a number of ways. But, the foundation of what it provides, the ability for us to deliver this together to our customers is just unbelievably exciting, because what it will do to the industry as a whole and what it will enable for the industry as a whole is like nothing we’ve ever seen before.
Now, our approach is just out of page one of the Microsoft playbook. We remain very focused on delivering this through partners and with partners, as well as directly to our end customers who will run it within their own datacenters. We expect a lot of servers to run, to continue to run, in that environment. So, our approach that we’ve had for many years, in terms of working with you, doesn’t change. The specifics may change, of how we work and what we do, and what you do, that will evolve, and we’ll learn all of it together. We don’t know all the answers. We’re on this journey with you. It’s a journey where all of the stops along the way are not fully known. But, we’re committed to very much do this together with you.
I know a number of you are asking the question, why does Microsoft run Azure, why are you guys doing this? And I hope I’ve been clear about the reason we do it is that we need to do it to learn, to be able to build the software that ultimately will power this next generation. It’s my goal, simply put, to continue to deliver the server and tools to power the world. And in the cloud world I’ve got to do it myself.
But, you may say, gosh, Microsoft is trying to compete with me. Let me just put this all a little bit in perspective. We believe in the cloud. We believe in it in all of our hearts. And today there are roughly 30 million servers in the world. Nobody knows exactly, because we don’t know how fast they are retired. But, we think there’s about 30 million servers in the world. And I believe that in 10 years half of those servers could be cloud-based servers, running in the cloud.
Now, 15 million servers, think about that, 15 million servers. We do a little back-of-the-envelope analysis, and we come to the conclusion that it will cost somewhere north of 500 billion to run and operate 15 million servers. We spend a lot of money on datacenters, a lot of money. We’re not going to spend $500 billion a year running datacenters, there’s just no chance. We’ll spend a lot, you’ll spend a lot, our customers will spend a lot. This is a broad ecosystem where the cloud will impact the way we deliver software, but it’s not going to impact the relationship that we’ve built over the years. And I look forward, I really, really look forward, to working with you, to understand how we follow this journey, how we move forward on this journey together, how we can do that in a way that enables you to provide very unique differentiation to your customers that’s way above and beyond what we’re doing.
We’ll continue to build the platform, and we’re committed to doing it in a way that is even better suited to meet your needs than we’ve done in the years past. But, I’m very confident that as we see all of this emerge you’re going to find all sorts of ways, and you’re going to see a broad, broad set of opportunities open up for you to take this technology that is shifting the world, and really use it to transform the relationship you have with your customers, and then to profit from it. That’s certainly what I’m committed to do, and I’m committed to do it with you.
Thank you very much. (Applause.)
END