Red Hat's new CEO aims Linux at the cloud
Red Hat's new chief executive, Jim Whitehurst, has his eyes on the sky.
The former Delta Airlines chief operating officer, who took the reins of the most established open-source software company from Matthew Szulik in January, names cloud computing as a top priority. Loosely speaking, the term refers to computing services available to anyone online rather than custom data centers isolated within corporate confines, but it also dovetails with the general idea of computing services running at massive scale on a more flexible infrastructure.
"The clouds will all run Linux," Whitehurst said in an interview.
Red Hat's new CEO, Jim Whitehurst
(Credit: Red Hat)Being Red Hat's CEO is always a balancing act. On the one hand you have the sometimes philosophically fervid open-source software community, volunteers and professionals who collectively produce the software Red Hat packages, tests, tunes, sells, and supports. On the other are the much more pragmatic customers who just want their technology to work. Red Hat must be friends to both camps--neither a parasite sponging off the hard work of others, nor a useless middleman selling what can be downloaded for free.
Whitehurst, who long has used Linux himself, discussed these and other subjects shortly before the LinuxWorld Conference and Expo, which begins August 7.
Q: What's your biggest surprise since starting at Red Hat?
Whitehurst: I think I finally get the joke. I was a senior exec, and like every other senior exec I had a huge IT budget. Mine was as large as Red Hat's revenues last year. You sit there and say, "Why are my IT costs going up, but I'm getting less and less functionality?" Every IT professional says the same thing: my lights-on costs are going up. But wait a minute! I bought a laptop, and it cost me half as much as it did three years ago, and my costs are going up? I get the joke now.
If you look at the S&P 500, seven of the top twenty companies are tech, and other than Google, they're not high-growth. But they're just printing money because switching costs are so high. There's this incredible amount of residual goodwill to Red Hat because we're seen as an alternative to that. Oracle announced a 20-something percent price increase just as the economy starts heading south. How can you do that unless you're pretty sure nobody can switch? High switching costs led to infrastructure cost creep. Once you get hooked, you can't get off.
I recognize Red Hat's a prominent alternative to incumbent players, but Red Hat's been around for awhile now, and it's not easy to get off Red Hat. It might be easier to get off RHEL than say, AIX from IBM, but...
Whitehurst: It's very simple. You can stop paying us.
But the switching cost is still there.
Whitehurst: You can stop paying us and keep using the same bits (the software). That's the point. The bits are free. You no longer get the support, but you can get support elsewhere. If you don't think we're adding value that year, you can stop paying us and keep the bits, so we've got to add value with service and support.
What if you want to switch to Novell's Suse Linux Enterprise?
Whitehurst: You'd have to ask different customers. A lot threaten us in contract negotiations over price. We luckily don't lose a lot that way. If you look at IDC numbers, there's about as much unpaid Red Hat as there is paid Red Hat out there, if you look at unpaid RHEL, Centos, and Fedora. Clearly people value the functionality, but a lot of people don't pay us for it.
What about all these high-growth companies with humongous scale-out infrastructure, like Google or Amazon? Does it concern you that these companies are able to use Linux for free?
Whitehurst: Amazon is very much a paying customer. Google is the very rare exception. As a company gets more sophisticated, one can argue the value of the support is less, but as companies get more sophisticated, the importance of the thing we provide goes up. So for instance, if Amazon wants to get something upstream into the (Linux) kernel because they need some functionality for EC2 (the Elastic Compute Cloud Web service), who can get it upstream? We can.
Today, we have Uli Drepper meeting with a bunch of major customers and invited Intel to talk about power management in chips, to talk about the next generation and what they need and what we need and feedback to Intel. If you're not a Red Hat customer, you're not there. That leadership in where the kernel is going is very important to most customers. It's much more smaller Web sites that decide not to pay.
Do the smaller customers pay when they get bigger?
Whitehurst: We see very little fee-to-free. We see quite a bit of free-to-fee, when customers get bigger, wake up, and say, "We probably need that support and certifications."
There's a lot of movement toward building large-scale Web infrastructure. Your product plays more to one server running the database here, one server running the app server there. What do you think about getting higher into the stack for enabling a very large, coordinated, distributed infrastructure?
Whitehurst: Like the Amazon EC2 cloud running on RHEL?
But what about the management, not just providing 1,000 individual operating systems?
Whitehurst: Most of the things we're doing now are all building blocks around effectively running grids or clouds. Whether those happen internally or externally, I don't know, but that's clearly the next-generation infrastructure. So with our virtualization strategy, which we call Linux Automation, we basically say applications certified once run anywhere--on bare metal, a virtual instance, or on the EC2 cloud or any other cloud running RHEL. If you look at the products we have coming out this year, Red Hat MRG for messaging and grid, it's all about the grid. We created a whole new business unit around management. IPA to start putting together the security. We worked with one large customer who live-migrated things from their own data center to the cloud and back.
If someone wants to move something from inside their firewall to outside, they need to feel confident it will run. If you're running it on Joe Bob's Cloud Linux, does it feel like it's certified? There is clear value to having a consistent Linux that's certified across those platforms. So we built the infrastructure to do that. It doesn't surprise me Amazon went with us. Anybody can run almost anything that runs on the data center on it, on it, it's certified, and they can call and get tech support.
What about management tools, though?
Whitehurst: We can live-migrate the things (moving running applications from one server to another). A lot of the tools are there. We're working on it, though.
But you can automate it--if I need 18 more instances on the cloud I can turn them on, then turn them back off?
Whitehurst: Absolutely. Our user interfaces are not as good as VMware's, but we have pretty extraordinary functionality we haven't touted as much as we should. (Amazon has) the single largest instance of virtualization out there. It's all our virtualization. They're moving workloads around all the time.
One of the interesting dynamics in the open-source world is the constant back and forth with Microsoft. Sometimes they're disparaging, and sometimes they're embracing. How's the current state of affairs with Redmond right now?
Whitehurst: The good news is I'm the new guy, so there's no personal baggage there. Clearly they're our single largest competitor, so we battle on a lot of things and have a lot of different opinions of things.
Have you seen any benefits from their interoperability announcements that came out of antitrust actions?
Whitehurst: My understanding is most of it is still vaporware. It's things they will do. Some of the interoperability things--they said they'd promise not to enforce (intellectual property rights) or you can use these APIs (application programming interface) as long as you're not selling. They're trying to relegate open source to a very narrow hobbyist niche. That's a bit problematic. But that said, we welcome a little regulatory oversight there and also welcome good hard competition.
Well, speaking of competition, how about Canonical. They're funding Ubuntu development aggressively. Do you see them in customer bids?
Whitehurst: I haven't heard of them in any customer bid. There are a couple issues there. They're primarily a desktop provider. We're primarily a server provider. Those are pretty different skill sets. I welcome Ubuntu everywhere. Let them get desktops all over the place. I think that's good for Linux and good for the ecosystem. It's just they have a single set of bits. They don't have a dual model with community and enterprise bits. I don't see how anybody's going feel real comfortable, without the certifications and support and enterprise nature of those bits, feeling comfortable running those in a mission-critical environment.
I talk to them every now and again, and certification is more on their to-do list than on their done list.
Whitehurst: It's the hard stuff. It's not real sexy stuff.
I draw a clear distinction between where we create value and where we extract value. You can't get too enamored with one or the other. The airline industry creates a ton of value, but it's never figured out how to extract any of it for itself. It's great for society, but never captured any money for itself. With our model, we create value by working with the community to develop really good software. We extract value by making open-source consumable by the enterprise.
Open source is iterative innovation--thousands of small projects going on out in the community. Iterative innovation is what makes open-source so powerful, and it is a disaster if you're running a data center. Every two years, we set aside the major bits, test them, performance-tune them for the major applications, etc. Making it consumable is where we're able to extract value.
The airline industry is charging for checked bags now. Does this feel like sunny optimism compared to your last job?
Whitehurst: Yeah. When you had an operational mistake you killed people. When you had a misstep you could literally liquidate the company. This seems a little simpler. Obviously we have our challenges, but let's just say I'll take a good business model over a bad business model any day of the week. We figure out how we add value in a differentiated way and extract it in a defensible way.
One question on extension is on getting subscription revenue beyond the operating system: How's that working, for example with JBoss (Java server software)?
Whitehurst: We have said that JBoss is growing twice as fast or more than our core software was in the first quarter, and we're targeting that for the entire year. It's looking good. We just announced our SOA Suite. It far exceeded our expectations.
This is where we stick to our knitting. Open-source, with our model, works best where certification, support, mission-critical (reliability) matter. That's where middleware matters, because it has similar sets of characteristics (to operating systems). Virtualization is another one. Those are the areas you'll see us focused on. It's where we can add value and make money.
So will you be sticking with the
Whitehurst: Yes. I want to be clear. On the exchange, I'm frankly not a big fan. We really have de-emphasized the actual exchange. The actual Red Hat Exchange program is extremely important because I do think we add a lot of value (at lower levels), but as customers want more solutions, we need to be working to ensure we have certified stack solutions--either appliances or certified stacks. So our RHX partnerships are extremely important there.
So that has evolved into a standard partner certification play, and maybe some cross-selling agreements?
Whitehurst: Yeah. What we found is a lot of people shopped on the exchange then went directly to the vendor to buy, which makes sense. So we've turned it more into a catalog. What we're working on hard is the appliances and certified stacks to make the stuff more consumable.
Red Hat has often been an advocate of open-source software, pushing the philosophical envelope, not just the business envelope. You're from the business mold, not the open-source advocacy mold. How do you see Red Hat's role now in leading the charge and waving the flag?
Whitehurst: I joined because I believe in the mission. I've been a Fedora user for years at home. Economically, it is a fundamentally better way to develop software. One thing I love about the company is it's a great confluence: when we do well as a company, we do good for the communities of use and societies around us. The last thing that wasn't open-source was Red Hat Network, and we just open-sourced it. We're all in. We are and should be leaders in open source, and that's the right decision for our shareholders as well.
Stephen Shankland writes about a wide range of technology and products, but has a particular focus on browsers and digital photography. He joined CNET News in 1998 and since then also has covered Google, Yahoo, servers, supercomputing, Linux and open-source software, and science. E-mail Stephen, or follow him on Twitter at http://www.twitter.com/stshank. 



"Cloud computing".....I guess this is the "next big thing" in computing now. It should definitely be more fun to watch than other developments in computing.
I never realized that Cisco or Oracle ever competed on the desktop. ;)
Gotta admit though - RedHat found their revenue in spades at the server level, so who can blame 'em for growing in that direction?
@PEdlund: You're joking, right? If you worked at the enterprise level at any position of actual responsibility, you'd know right away that there is no such thing as a one-size-fits-all solution, period. Each enterprise-sized group (or entire corp if applicable) has their own defined goals, and more often than not build their own solutions with what's at hand.
Open-Source software is the absolute best at letting them do that... by being highly flexible and extensible, but at the same time being highly reliable. I can build my own solutions with my own in-house programming team - for cheaper than it would cost to license semi-sort-of-fits-but-not-quite "solutions" being hawked by a third-party vendor (see also WebLogic, SharePoint Server, et al). If I can do my own support, why should I be stuck on a vendor treadmill? If I already bought the software and I'm building my own solutions from it, why the hell should I have to pay ongoing license fees (e.g. CAL's)? Unless I'm specifically renting an ongoing service or a physically tangible product, I should only have to pay for it once and be done with paying for it without good reason.
As for the farcical: "Everyone assumes that someone else will do the work for them", you couldn't be more wrong. Out here in the large enterprise, we assume that we will do our own work - we just want a nice and flexible set of building blocks from which to do it, so we don't waste time re-inventing the wheel. My employer has literally several hundred programmers whose collective job it is to tailor in-house solutions and to make them fit our needs.
The really cool part of FOSS is, more often than not a lot of these solutions and parts thereof make their way back into the world at large - and are licensed so that nobody can steal them, but have to share them. You're under no obligation to do so at all if you're not distributing the solutions or modifications... yet strangely enough, it happens all the time.
Maybe someday you'll realize that... or maybe not.
BTW, your argument has no merit because of (at least) the following:
1) code to the LSB, and distro choice no longer matters. It also doesn't matter which distro you use if you're writing your solutions in Java, Perl, Python... and if it's web-based in the first place, who cares what's behind it distro-wise? Of course, if you actually knew what you were talking about in this biz (instead of creating sock-puppet accounts all day long, "CTO_Dude") you might have known that.
10 times the resources? Really? Because I'm able to stack services onto a typical Linux server that would bog down the same hardware running Windows Server 2k3 (and judging from testing, 2k8 as well). There's a vast difference between writing solid code to start with, and throwing hardware and bandwidth at the results of a crap programming language.
Roadmaps are easy... building and keeping to one that doesn't boil down to: "pay Microsoft and some third-party vendor tons of money to get an ill-fitting solution that we'll have to replace in 5 years anyway" is a bit harder to do. OTOH, it's worth the extra planning when you do it right, and the budgets seem to have more room in them when I'm done.
Ever hear of designing an elephant by committee? That's exactly the problem.
Cloud Computing is something that online providers like Apple, Google, Microsoft and even Amazon (who I believe will be entering the fray more seriously) are going to have the advantage. They only have to satsify the customers, not ego-driven geek wannabe's.
What Whitehurst meant was that Apple, Google, Microsoft, Amazon are all going to do the cloud computing. Redhat is not going to compete their. But they will all need Linux at the base. Some will do it on their own like Google, others like Amazon will buy support from Redhat. Apple and MS both have their own OS. But those will not be the only players, there will be a lot of small players and those will have a high probability of chosing Linux and Redhat expects to compete better than other Linux vendors.
First, AFAIK, Apple hasn't much Enterprise level framework for Grid/Cloud computing, then Google's stack is heavily based on opensource (which isn't synonym of availability for everyone at no cost) and finally, if you actually read the article, you would have learned that Amazon cloud is RedHat work.
Next time, when you don't know a topic, just refrain from commenting as it can become embarrassing for you.
Linux is like a liquid... it seeps into ever nook and cranny that gravity takes it to.
I have seen no forward movement in the use of Ubuntu server in business. If you look at who contributes to the kernel (and various other large infrastructure related software pieces) Red Hat has consistently shown up on top. Check out Greg Kroah Hartman's fairly recent presentation on the Linux kernel development process for the answer on "how much does Canonical contribute back to the community" question (See: http://www.youtube.com/watch?v=L2SED6sewRw).
Red Hat makes a "corporate desktop" and I disagree with your comments regarding the quality of their desktop distribution.
Regarding businesses switching from Windows to Linux, it is quite common.
Ubuntu might "owe" the Linux desktop but they get close to zero money out of it. Red Hat was in that position about 8-10 years ago but realized that it would be near impossible to make money from that. Not much has changed since then (with regards to profitability of Linux on the desktop) and it is unclear if and when Canonical will have a workable business plan anytime soon. I guess that is fine as long as Mr. Shuttlesworth continues to operate is as a charity organization. It would be nice if Ubuntu were giving their changes back to upstream but that doesn't appear to be happening. From an end user's perspective Ubuntu might be good, but in the long run, as long as they continue to ignore upstream, not a whole lot is going to change.
Red Hat may not be targeting the desktop market directly... but they are funding a significant amount of work on desktop related technologies... indirectly through the Fedora project and directly by being the lead developers on a number of desktop oriented projects.
If and when the Linux desktop becomes financially viable (and it isn't for Ubuntu now), rest assured that Red Hat will be interested. In the mean time they are moving in that direction even though they don't have a current product line that addresses the desktop.
- by ashyanbhog August 10, 2008 12:35 AM PDT
- Ubuntu maybe the latest cool kid in the Linux landscape, but only Novell comes close to Redhat when it comes to contributing to the community. Its sad to see people bash Redhat without reason.
- Reply to this comment
-
(13 Comments)Remember, to date, they are the only successful growing business in open-source.