'Dogfooding' determines the fate of new Microsoft products
Seattle Times technology reporter
Running the computer network at a big business is a hard job. It's even harder if your company is the most scrutinized technology business in the world.
But it really gets rolling when you're expected to run test versions of software that don't have all their bugs worked out — "eating your own dogfood," in Microsoft parlance.
That's the job of Rick Devenuti, vice president and chief information officer. Devenuti, 45, grew up in Federal Way, the son of a Boeing engineer. He worked his way through the University of Washington selling jewelry at Friedlander's in Southcenter in the evenings and on weekends.
After graduating in 1983, he worked for a predecessor of Deloitte & Touche, doing audits for a growing software company then in Bellevue.
Devenuti was on the accounting team that helped take Microsoft public. A year later, in 1987, he joined its finance department.
"When the opportunity came up, it was the place I wanted to be," he said.
Today Devenuti leads 3,000 people running Microsoft's internal networks. That includes 7,000 servers worldwide supporting more than 60,000 employees, interns and vendors. Outside his realm are MSN, Microsoft.com and Xbox Live.
Devenuti's division is taking on a greater role developing and selling software as Microsoft pushes further into the enterprise software market. The company has always tested its products internally, but Devenuti formalized the system to the point that his division must sign off on products before they go on sale.
Microsoft is also using its network to showcase products and using Devenuti as a pitchman. When Windows Server 2003 launched in April, Microsoft was running it on 2,500 servers.
Expect to hear more from Devenuti later this year when the company launches new versions of Office and Exchange.
Here's an edited transcript of a conversation with Devenuti about his role at Microsoft, tips for information-technology managers and, of course, dogfood.
Q. Let's talk about dogfooding.
Devenuti: One of the primary objectives of this organization is to be Microsoft's first and best customer, so we run the environment on beta software. We made that decision about three-and-a-half years ago because we thought the best way to add value to the company would be to make sure our products are ready to ship to enterprises when they ship. ... We said there's no way the product can be tested the way we can test it — putting it in our environment — with 50,000 users in 400 sites over all kinds of LAN connections — and really roll the stuff out.
Q. Why is there a PC running Red Hat Linux in the corner of your office?
Devenuti: To keep an eye on competition. If you don't spend time with it, if you don't learn, if you put your head in the sand and pretend it doesn't exist, we're never going to make better products.
Q. What do you think about the software?
Devenuti: I think it looks a lot like Windows.
Q. When do you expect to start running the upcoming version of the SQL Server database (code-named Yukon)?
Devenuti: Yukon is just starting to get into the data centers. Today SQL 2000 is the primary database. I know there are servers that are testing now. I haven't seen a plan for when we start to deploy it as our core.
The way we go through the process is very formal. I don't want to give you the impression it's a test lab; it is actually a disciplined rollout of technology. The way we do that is to sit down with the product groups and say, "OK, what is it you're trying to do?"
Q. How do you work with product groups?
Devenuti: What we actually have is an agreement: They can't ship the product until we do what we agreed we would do. If we run into problems, we stop and go backwards. We need the fix. As we get it, we roll again.
We call it shared goals, (something) we're absolutely joined on the hip at. As we get that done and we sign off the product and say it works for us as an enterprise, they do the rest of their work, then they can ship the product.
Q. At what point do you start trying out new products in your network?
Devenuti: The cycle of every product's a little different. Beta 1, 2 and sometimes 3; release candidates 1, 2, sometimes 3. We generally start to move it at Beta 1. That's in the closed environment. We tend to move into large- scale production for the company at RC1 (release candidate 1).
Q. How has this affected the relationship with product groups?
Devenuti: I think the product groups each had a different view about the value that was added from the IT organization. It's now just an accepted process, as part of product release, and it's fundamentally changed our relationships with the product groups.
Q. Did the more formal approach to dogfooding coincide with the company's increased emphasis on the enterprise?
Devenuti: Absolutely. It just makes sense as we move into the enterprise that if we couldn't say we run it, it wouldn't make sense. Its very different than Office — everything has to connect, it has to work with different products from different vendors that are naturally in the data center, you've got to be able to monitor it, you've got to be able to run it remotely ... . It's just very different than a desktop.
And so while you could use test labs and beta users for desktop things, you can't do that in the enterprise. You need to have an infrastructure, because it's all about integration.
Q. Have you reduced the number of servers at Microsoft?
Devenuti: It has gone down in specific cases as we've been able to consolidate. On the other hand, the company continues to grow, we continue to add new applications, we continue to add more customers to our Web sites, which require more servers — front-end servers and clusters, to meet the traffic needs.
Q. Your boss, John Connors, had your job before he became chief financial officer. Is that your goal?
Devenuti: You know, I'm really happy with the job I have right now. I think I have one of the great jobs at Microsoft. There are some things about this job I don't know where else you get — I get to work with an incredibly technical group, I get to work with a lot of customers — because we're early in what we do and customers want to know about it.
I think John's role, the role as CFO today, is a tough job. I'm not sure he's having as much fun as I am, based on what's happened in the economic environment over the past two years, in terms of the reporting requirements — Sarbanes-Oxley for example. I don't think that's as fun as rolling out the next version of Exchange.
Q. How early do you get involved in scoping products?
Devenuti: We are involved in the very early stages. As we finish one product, we move right onto the next with the product groups, so we're deeply aligned with the Longhorn (next version of Windows) team right now — what features are they thinking about, what feedback do they need, what is it in the current environment that we'd love to continue to work with, how do we make it easier to monitor, how do we make it easier to move bits to servers, how do we make it easier to reduce the number of people we need to run the environment.
Q. How close are you to five nines (99.999 percent network uptime) and how is that affected by dogfooding?
Devenuti: I think five nines is mostly a ridiculous concept. Five nines is five minutes a year downtime. The cost to get there in any environment is very high. You just have to think about what you do.
For Microsoft it doesn't make sense for us to spend the money to get to five nines. For some of our customers, it does. For a bank, the ATM machine better be up.
You can get that. First of all you need great software, you need great processes, you need clustering, you need redundancy — it's doable. For virtually all of our processes to be five nines doesn't make business financial sense. So we run most of our systems at better than three nines to four nines.
Q. Do you have any tips for IT managers besides "buy Microsoft stuff"?
Devenuti: We worry about the same things — worry about security, worry about total cost of ownership and what our budget is. We worry about our people: Do we have the right ones, are we training them correctly, are we motivating them? We worry about issues at an enterprise.
So we constantly focus on three things. Technology. What's the right technology, how do we pick it, how do we think about it?
You've got to have great technology partners, you've got to have great partners. For us it's partners. If it's not Microsoft, pieces in the network system, pieces in backup, Siebel or SAP, you've got to pick great technology.
You have to have great people who are really motivated and really educated about what they are doing. And you have to make sure you build processes that allow those people to seamlessly and repetitively implement it.
Brier Dudley: 206-515-5687 or firstname.lastname@example.org
Copyright © 2003 The Seattle Times Company