There are sort of two trends in the operating system space right now. At the high end, you have Unix systems, which are becoming more like mainframe systems. The vendors' goal there is to make the operating systems more resilient so they can adapt to hardware failure and to shave downtime to a minimum, because that's really what's needed for the highest end, the most critical workloads. The key is to become resilient to failure, which means that the operating system has to detect when there is a hardware failure, because hardware failure is inevitable.
The newest operating systems can adapt by detecting when a hardware failure occurs and then moving the application away from the area of failure so that the application keeps running. At the end of the day, that's what the operating system does. It's an intermediary between the hardware and the application. It will try to protect the application at all costs. Mainframes have been able to do this for awhile, but you're starting to see some of those mainframe features now trickle down into Unix. What's happening at the low end?
At the low end, which I'll call Linux and Windows, vendors also want to be able to do those things. The developers of those systems want to make the systems more adaptable to failure, but they have a lot farther to go. Anything that Unix does will ultimately be done in Linux and Windows; it's just going to take a few years to get down there.
First of all, you have a different set of requirements there because you have a lot of small and medium business users and desktop users who don't necessarily care about having that much uptime. They have more flexibility in terms of being able to reboot the server or something like that. They don't really need those super high-end features that you have in Unix systems.
The other issue is that [low-end operating systems] are designed for industry standard hardware, for Intel-type systems. And that makes it a little bit difficult to implement some of these features because you sort of have to wait for Intel to do the things that are needed to support the functions. Can you give an example of one of these functions?
A good example is virtualization [or using software to emulate multiple instances of hardware]. The use of Intel hasn't yet delivered that capability for their processors, so virtualization has to be done in software. On the high end, with RISC systems, the designers have the ability to design the hardware and the operating system together. They work sort of hand in hand to come up with a total system that can do everything needed to maintain uptime. We'll eventually get those capabilities with Linux. Windows is going to take longer. Is Microsoft focused on this issue?
Microsoft in particular is focused not just on functional capabilities, but also on ease of use. That's always sort of been their differentiator. You know where Linux developers really concentrate on functionality and making sure that you have high degrees of reliability and security and uptime and so on, ease of use tends to be an afterthought. Microsoft really tends to tackle ease of use from the ground up. They're also trying to make Windows more resilient to failure, but they're mostly thinking about the users' difficulties in managing information. How will Microsoft approach the problem of managing information in its upcoming Longhorn operating system?
WinFS [the code name for the storage platform in Windows Longhorn] is being designed to make it easier for users to manage their information in a very dramatic way, basically by turning the file system into a kind of database that can be searched. They've been working on this for a long time and now they're saying that it will not be available in the first versions of Longhorn. It's not going to be in there right out of the chute. When can users expect to see the new Linux 2.6 kernel in action?
Linux 2.6 is already available from SuSE. But Red Hat has not shipped its 2.6 kernel. That's going to be coming up in early 2005 and will be called Red Hat Enterprise Linux 4. What are the biggest developments in the Unix space in terms of product news?
HP-UX will be shipping an update to HP-UX. That's the big development in the Unix space. They are adding virtualization capabilities to it and they're going to be extending their workload management capabilities that they have in Workload Manager. And they're going to bring out native virtual machine technology for HP-UX. IBM and Sun both shipped updates to their Unix OSes [in 2004]. This year you'll also see [Sun Microsystems'] Solaris 10 really enter the mainstream. Sun Microsystems had been experiencing some serious financial woes in recent years. How are they dealing with those problems?
Well, everyone wants to know that. [Laughs] They are transforming their business model from that of a supplier of SPARC systems that happen to run Solaris to more of a solutions provider where you have a choice of platforms on which to host your workloads. They're switching to industry standard x86 platforms. This is really a push for them to become a preferred supplier of operating systems on the Intel/AMD/x86 platform with Solaris. They have their own AMD systems that they're selling.