One of the reasons we've made these statements this week is that we want to make enterprises understand that this is not just a marketing campaign around Linux. Linux runs well, it saves you on hardware costs, and Oracle runs well on Linux -- you should move there as well.
Internally, we also want to save money. We started 18 months ago moving our development organization to Linux. Our applications division was first -- human resources, CRM, manufacturing. That group [5,000 programmers] finished its move last October to Linux. Whatever e-business application development we do is done on Linux. Do customers care?
This is a huge thing. Our customers look at us to see what base platform we develop on. Applications was the easiest stack to move first because it does not heavily rely on the operating system. It runs on top of the database, and it was easiest having the servers they use for development to move to Linux. What division is moving next internally?
The second step is our product development for our databases, application servers and collaboration suites to Linux. At the beginning of the year, we started getting the hardware in and doing the rollout. We expect to have those 4,000 developers switched over by the end of the year.
Internally for us, Intel hardware is cheap compared to what you would get from a proprietary [Unix] vendor. From a development point of view, systems are so much faster. For example, if a developer has to write a patch and has a tight deadline to get the feature checked in, [we're noticing] it takes less time with Linux. We're seeing a reduction in many cases from 12 hours to four hours. Our feature queue is running two to three times faster on Linux.
This is a long-term commitment and a really important one. This is a clear message to our customers that Linux is going to be there for a long time. We've made an investment, and we're establishing a comfort level for customers. Gartner Inc. has some favorable database-on-Linux numbers this week. What's the key takeaway in your opinion for an enterprise looking at those numbers?
It shows that customers believe in what we're doing on Linux, and that there's a strong presence of Linux on the database. The numbers also show that it's not just new customers buying Linux. There's clear evidence that existing customers are switching to Linux. Those that have made the switch are happy because they're saving money on hardware costs. What's driving Linux deeper into the enterprise right now? Kernel improvements or support from Oracle, IBM, HP and others?
It's a combination. Before we started Unbreakable, Linux was not ready to be run on big database servers. It didn't have the features necessary to do so. It's been a combination of Oracle, IBM and HP support, and all of us telling Red Hat and SuSE that we need this [Linux] in the operating system before we could tell our customers it was a good choice.
In 2002, Oracle 9I r2 supported Linux; we worked with vendors and backed it up. The most important thing was that we said we'd back it up. We'd support it if you were a Red Hat customer running Oracle, just call Oracle and pay for database support and we'd give you free Linux OS support.
It took a few months for things to kick in, but since 2003, a lot of our customers have gone live with Linux and switched the entire server base of their companies over. How has the switch to Linux been embraced internally at Oracle?
A bunch of our developers already use Linux at home and were familiar with it. There were some who really didn't want to switch because they'd have to relearn a technology. But, Unix is Linux with the same complier tools; maybe there are a few new shell prompts they had to learn. Once we did move, a lot of them were saying Linux was a lot faster. They were comforted by the performance improvements. Is the hype over the 2.6 kernel real?
2.6 will be in the next distributions [from Red Hat and SuSE], and it will solve all the problems we still have. As for Oracle customers, a lot of the improvements to the stack that will be made, we already ship. We don't need clustering support, for example.