Are enough CIOs and enterprise decision makers thinking about using Linux and open-source software from an architectural perspective?
The only companies thinking about it that way are the ones being successful. I am not seeing companies able to change [the] status quo or retrofit infrastructure purely at a technology level. Without having the architectural vision, things are constantly mired down in topics of speeds and feeds and discounts and SLAs and features in the next release. It ultimately becomes an environment of no decision.
But when people come at it from the architectural perspective, they realize they can make the structural change that will not only cut short-term costs but enhance long-term opportunities. Then make the commitment. If that commitment requires a little bit of extra effort, they are going to see to it that extra effort is made.
The happy thing about the migrations to open-source that we've sponsored and supported is that they have had positive ROIs in 90-day and 180-day time frames. If you read the trade press three years ago, [you know] you'd get laughed out of a business meeting if you'd asked for a year, let alone three- to five-[year ROI]. So the fact that we've been able to hit our 90-day targets means there's no P&L exposure and there's 100% upside; that makes it go forward. The good news is that it doesn't require a lot of faith to go forward, but it does require an act of commitment, and that's where architecture comes in. Is there a mindset change that still has to happen, one that Linux and open-source advocates still need to evangelize?
I'll tell you about the funniest mindset I've encountered in the last week. Unix-to-Linux [migration] has now reached a point where the majority of customers I talk to have already done the migrations in their mind. I'm talking to people at very large proprietary Unix deployments and, when I talk about the cost savings, benefits and opportunities of Unix-to-Linux migration, the responses they're giving me is, 'We've already done that. What's next?' I ask them, 'Are you still running Unix?' They respond, 'Well, we've done it in our budget forecasts and in our models; we don't need to be convinced of that.' It's kind of like people who are going to quit smoking. They've all quit smoking in their mind.
Unix-to-Linux migrations have now become accepted. Now the only question is, how do you take action? That again is where the architectural thinking comes into place. It's not just about saving money. Yes, there's a very strong agenda for companies to be more responsible about how they manage their financial resources, but it's all about how to open up the longer-term future of the company, how to make sure the company can get IT deployments to happen more quickly and to happen at a higher level of quality, be more manageable. People do not have experience in any of those dimensions in the proprietary Unix world. It does take a little more work than, 'Yeah, Unix to Linux, let's save a lot of money.' It requires them to realize 'this can actually change how I do business.' Is there a turning point people need to see?
I think the No. 1 turning point people see is when they finally do have the staff meeting and the executive asks the staff, 'Who wants to run with this ball?' The turning point happens when someone says, 'We actually already have it deployed in these areas. How about if we just get the official blessing?'
The second dominant turning point is for companies who have leased instead of bought their infrastructure. If you've got 500 machines coming up on lease and it is a seamless transition, this is an opportunity to have a huge financial impact.
The third thing is when the status quo system suffers a failure. One of our financial services clients had an options-crossing engine, an 8-way proprietary Unix machine, and the capacity demand on that machine had reached 100%. When it reaches 100%, a bad thing happens, your response time goes down. What happened was, this particular system experienced an asymptotic increase in trouble tickets. Every trade was a problem.
More than 60 days ahead of schedule, they replaced this 8-way Unix system with a 2-way Egenera blade running Red Hat Linux. That day that it went live, Feb. 14, 2002, they were able to cross five times the trading volume they had ever recorded, which meant there was a lot of stuff they weren't doing [beforehand]. They were missing opportunities. Not only did they increase their actual delivered capacity by 5x, but the load factor on the machine was 16%, and that's going from an 8-way to a 2-way. So, these guys who are saying we're not going to look at Linux until it can support a 64-way machine, the question you have to ask is, 'What are these guys smoking?'
The [fourth] point is that the production system fails, and that's the time when you say 'the heck with it, I'm going to do what everybody else is doing.' The SCO suit. I'm sure folks at Red Hat, SuSE and other Linux distributions have played out the what-if scenarios. What if they win?
There are enough crazy lawsuits in the world. There's a lot of things we could worry about. What if somebody trips on our steps? The what-if scenario is pretty broad. I would say that this lawsuit sort of ranks with landing a plane on the White House lawn. Could somebody do it? Yeah. Is that really dangerous? Well, somebody made the comment that, if there is any merit to this lawsuit, IBM would darken the skies with lawyers. I think that sort of answers the question. UnitedLinux. Does Red Hat have a future there ever? Does UnitedLinux need Red Hat more than Red Hat needs UnitedLinux?
At a Wall Street roundtable last year, Red Hat and representatives from SuSE were invited to present Red Hat Advanced Server versus UnitedLinux. In July, there was no formally released UnitedLinux. There was this concept of 'someday we'll support the things that Red Hat supports, and someday we'll have the ISV support that Red Hat does, and someday we'll have the hardware support.' But, in the meantime, download our beta and go with our model.
At that point in time, I made the point that Red Hat Advanced Server is 100% open-source, that if UnitedLinux wanted to at least shorten the amount of time it would take to demonstrate functional versions of Oracle running on the platform, because of trademark law, they couldn't call it Red Hat Advanced Server, but they were welcome to call it UnitedLinux 1.0 if they wanted to. If they did that, at least they would have software that was known to work.
There's a second half to the story, and that is having the relationship with these ISVs that makes enterprise support actually effective. Certainly, UnitedLinux has a close, some would say too close, relationship to IBM. The PeopleSoft announcement citing Red Hat as their primary platform, this is the enterprise logic we see at all ISVs. Given Red Hat's market share and given ISVs' desire to make a profit, it makes a heck of a lot more sense if you can get access to a majority of the market for one single investment and, if you have to double that investment, it only gives you a small percentage of additional market; it doesn't make sense to make that second investment until that small percentage is actually a large number. The happy day when the Linux market is $10 billion or $100 billion, spending the cost of qualifying on a second Linux platform would have a positive ROI. But today, these ISVs are showing positive ROIs by making one investment. But the numbers don't support making a second investment.Opteron. How big a deal is this for Linux?
Opteron is interesting because it adds a new 64-bit option. There are a lot of people who think the seamless 32-bit to 64-bit migration is a seamless facility. There are a lot of people who think, 'I won't look at Opteron until it is shipped in enterprise-class skin.' IBM has announced [its] intention to ship Opteron in maybe Q4. That may mean some people won't actually evaluate it until it shows up in black iron. But, regardless, at the launch announcement, they took our content-acceleration engine, which holds all these Web server records, and -- lo and behold -- on Opteron, we further advance the performance metrics to unbelievable levels. One of the great things about Linux is that it's not tied to a single vendor. So when innovation comes up, whether it's at the silicon level, the motherboard level or the mainframe level, Linux can accommodate it as a part of a larger platform, which is a platform of choice.