Bill Laing, corporate vice president of Microsoft’s Windows Server Division, recently sat down with Michael Otey, technical director of Windows IT Pro, and Michele Crockett, editorial director, to discuss the release of Windows Server 2008 R2. As Laing pointed out, this is the first server release to be timed with a client release—Windows 7. Laing talked about the scenario-focused development process that produced both products and the tight integration with third-party partners. Although Windows Server 2008 R2 doesn’t lack for cool new features, its most riveting qualities—such as Core Parking—are those that save resources, time, and aggravation for IT pros. Here Laing tells how listening closely to customers and partners yielded a product that helps businesses operate more efficiently in a tight economy.
Michele Crockett: What have you heard from the launch events that people are most interested in and excited about?
Bill Laing: We had a great reception to V1 of Hyper-V—particularly Live Migration and cluster support. And when I think of server I think of two things—I broadly think of the attributes of the basic system: the core system. But I also think about it by workload—the file server workload, the Active Directory, file and print, web server—and I think of these as separate. Each workload offers some exciting things.
For example, Active Directory has the Recycle Bin, which seems like a fairly small feature. But when you talk about it you get spontaneous applause. When we were planning the release, we did a number of things to get input from different people. I split feedback into three categories: the voice of the customer, our business people, and the technology. We asked ourselves, "Is this the voice of the customer, the business, or the technology?" Moving into 64 bits or virtualization are technology drivers, and then you might turn it into some customer need as well.
I use what we call red zones. Every quarter we take all the calls to our support center (CSS), and they catalogue every call—how long it took and what area it was. I have them roll it up and give me a sort of inverse of the most calls, the longest ones, and I use that to prioritize things. And the calls aren't always about a system crash. One of the highest-level calls, from my point of view, was from customers who had deleted accounts accidently from the domain controller, or they had corruption and didn't have a backup. That was one of the things that drove \[the AD Recycle Bin\].
Michele Crockett: Are you getting a feel from people about adoption and how quickly they might move to R2?
Bill Laing: It does seem to be a little faster than 2008. I'm not quite sure why—I think they got used to 2008 and it feels less disruptive to them, and we try to do that with an R2 release.
Michele Crockett: And some companies are probably moving from 2003.
Bill Laing: Yes, as you get close to the end of the support cycle, people start to think about 2003 and upgrade plans. Virtualization also gives them a bridge to the past if they want, because they can move to that and still run old programs. And there's a lot of pressure on cost and power. A couple of customers I've talked to, who just by consolidating the number of machines and running in VMs, have gone from 15 machines to 3. Forget any fancy software we did on power saving—they're running 3 machines instead of 15!
Michele Crockett: Our readers talk about working with limited budgets. They ask what they should do first, what things really do need to go together. Do you have any insight into that, to address the IT pro who doesn't have a large budget to work with?
Bill Laing: Virtualization is one of the biggest and easiest wins, because you can reduce the number of machines. It's hard data to get, but if I had 50 machines, and still keep 50 VMs, how much less is my management effort? It's clearly less, but it's not like running 1 machine; you still have 50 OSs to run, but there are some savings there. We did a lot of power management in the new OS, and it's pretty dramatic—it exceeded my expectation of what we would do.
Until recently, idle servers typically consumed just about as much power as when they were fully utilized. We didn't do much power management, and there wasn't a lot on the chips. But we've driven that down dramatically to 40 percent to 50 percent power utilization when servers are idle.
Michael Otey: Core Parking is a cool feature because it's enabled automatically right out of the box. On one of the first systems I used, I opened Resource Manager and I could see that as the system is active, all the CPUs in it are active. As you stop doing things, the system automatically parks them and there's nothing you need to do to make that happen.
Bill Laing: It used to be, "Well I have all these cores, and I'll just run my workload across them." Now you say, "What's the least number of cores I can use to get through the work." It's the small changes like this, where people don't actually have to do anything, that get attention. We've also done a lot more work on PowerShell so people can automate a lot more.
Michele Crockett: That's a big deal with our audience—they love PowerShell.
Bill Laing: We have a blog and a site for PowerShell, and it's one of the most popular places that people come to. People like the idea that you can take a script that someone else wrote and perhaps refine it or reuse it, so there's another big time-saving opportunity.
Michael Otey: The new PowerShell Integrated Scripting Environment is really nice. That feature really helps people and administrators get started with PowerShell because it gives them a graphic, color-coded editor that makes it really easy to learn.
Michael Otey: R2 contains a lot of features. Can you give us some idea of the development scope? How many people worked on this and for how long?
Bill Laing: It's hard to estimate the number of people, because we share the work between the client and the core. There's a core OS team, a client team, a server team—and we tend to think of it as the base system and then the workload. But some team started probably about a two-year cycle of development. Some people started before 2008, and one of the things I've tried to drive is that it's much more important to deliver complete scenarios for end-users than do lots of features that you have to complete yourself, or all the pieces aren't there.
So we use a model called CMD, or customer-focused design, where we interviewed a lot of customers and partners (hardware partners are a key part of the release) and then tried to build the core things that customers told us were needed. And we very specifically wrote down everything in their words, not our words. Traditionally, we'd interview people and the employee would come back and say, "Well what they really want is this," and that would be the thing that the person interviewing them wanted to build, not what they really wanted. So we disciplined ourselves to write down what they said and we used that to guide us.
And the second part of that was rather than saying, "We'll ship when we have less than X number of bugs" (which would be a traditional way of doing it), we used guidelines called CTQs (credible to quality). So we might say we can support X many users in a VDI session doing this, or something around the speed of Live Migration will be this speed, rather than saying Live Migration ran and didn't crash in 24 hours. Every team had CTQs that we'd hold them to, which is a different model than we've used in the past.
We also have a lot of telemetry in the system, so we can measure the roles people have installed. People can opt in to provide data anonymously, and our beta site customers (we call them our TAP customers) actually enter additional information. TAP is our technology adoption partners, for customers who would take the airway release and give us feedback. There used to be people whose job it was to call them every week or two weeks and say "How many servers have you deployed, I have to report back to my boss." Now they're automatically enrolled and we get the telemetry and we can tell them "By the way, you're actually running 53 web servers" and they'd say, "Wow well we didn't know that." We measure what roles are installed, which tells us the coverage we got during testing. In the past you really just had to hope sometimes that you put the beta out. So I feel really good that we've got a good rhythm going with customers. We've built up that level of trust.
Michele Crockett: I think IT pros might still be used to waiting for the R2 version, even though I know Microsoft has tried to change the release rhythm. Do you think you've made any headway with people not waiting for R2?
Bill Laing: I think it takes a long time to rebuild that belief in you, and I feel like we've had a really good track record from 2003 onward—relatively predictable release seasons, good time frame, not waiting for the first Service Pack to come, so I think people are more comfortable with that.
We also work closely with our hardware vendors to make sure that they are systems-ready at the release. Sometimes that's their main interest, sometimes they have core features that they want to work on with us before the release. Typically, Intel, AMD, and the bigger OEMs have some feature they've invested in and want it to be exploited or made available through the release. I think we have 1,200 Windows Hardware Quality Lab certified servers for 2008 R2 already, which is much more than we've had before.
I'm very pleasantly surprised with how smooth the move to 64 bits has been. We made that decision about four years ago—I think we announced it at TechEd 2005—and said 2008 would be the last 32-bit, and we'd try to move people. As far as I can tell it has gone very smoothly—we're not seeing any hiccups, the drivers are all available. You might hear otherwise from your readers, but again our telemetry was telling us what we could see was happening with 2008.
Michele Crockett: What does R2 offer for people who are leaning more toward Linux, either users already on Linux or that have been a bit fickle?
Bill Laing: There's a couple of things. I always say to people: focus on delivering value. When you look at the total cost that people spend on IT, the actual procurement cost of the hardware and software is a pretty small percentage. I've seen numbers between 10 percent and 18 percent. So if you focus on price, that's interesting; but if you can help people with the other 85 percent to 90 percent, reduce their costs, and make their lives easier, that's better. I'll often say, "Let's figure out features that give IT pros back an hour of their day." That's actually worth more to them than the original cost.
And also, there's the fact that we have an integrated system—you can put a number of roles together. I think we still have a very strong directory, which is a big part of the infrastructure—we play well with the clients. So I believe that, ultimately, people make rational economic decisions over the long term. But it's like saying the stock market is rational over the long term.
Michael Otey: What are the most popular roles being deployed? What percentage of Windows systems are running multiple roles such as file server or active domain controllers?
Bill Laing: It varies depending on the size of the business. Enterprises tend to be more dedicated per role. Branch offices are a funny case because they behave like small businesses, but they are an enterprise deployment and they tend to put multiple roles together. So we've seen the Read Only Domain Controller plus File and Print Server, and maybe running an application locally in a branch or small office.
Similarly, they might have a web server because they have SharePoint. And at the high end, enterprises are more likely to say, "This is my web server, this is my directory server," and that's why we came up with the term "role." When we would interview people they'd say, "This is my…" and they'd point to the machine.
So we have a reasonable spread—it's the ones you might think of, such as file and print, web, that are all fairly created shares. Terminal Services is a popular deployment role as well. Virtualization is something else we're seeing that's changing the dynamic there a little bit: people are saying, "It's nice to keep things separate but still consolidate them on one machine." Because there's always, "If I change this for this role, are there any side effects to other roles," whereas you can put them in a separate VM and still get that benefit.
Michael Otey: In some ways virtualization could let you more easily have one server per role because then you have a virtual machine that's your web server, another one that's your Active Directory, and so on.
Bill Laing: Yes, and it's easier to do your upgrade or make some configurations if they're really only for a web server and not your domain controllers, for instance. It's interesting that with R2 the domain controller is very compatible, so we tested with earlier versions of Exchange, the latest version of Exchange, earlier versions of Active Directory, and we've learned that trying to change the infrastructure really slows down deployments. And you can say that you can really happily go from NT 4.0 to Windows 2000, but by the way you have to roll out this new domain structure. It's really hard to get the deployments then. And you can just do the features you want for Direct Access, and you can just put that part to support your Windows 7 clients, or put in a Branch Cache for a Windows 7 client.
I think the momentum around Windows 7 has helped. It's undoubtedly a very positive response from people. It really is a very exciting system whether you're an IT person or an end user—we're seeing a lot of excitement from them.
Michael Otey: R2 has an incredible number of features. If you were to pick, what would you say are the top three value-adds for your customers?
Bill Laing: In priority, it would be virtualization and Live Migration. I think the power management work we've done is a slow burner people will appreciate over time. And I like the more extensive PowerShell coverage.
Michele Crockett: What customer scenarios did not make it into this release that really broke your heart, or that you're looking forward to including in the next release?
Bill Laing: I don't think anything broke my heart—we're just starting to retrench and get feedback from people on what they want. I think we learned a lot about the process—we've improved our maturity and planning process. Really being connected to customers and partners is key—if you're not building something they want, maybe you're not going to be the king of the market. I think we feel very attuned to that. I'm really a big team player, focused on partners and customers. We have a pretty well-developed process for collating our feedback, as well as telemetry, and the competition is Linux and VMware. It's somewhat simpler than it used to be in some sense.